You are on page 1of 6

A Knowledge-Based Approach to Voltage and

Power Control in HV-MTDC Grids

Adedotun Agbemuko, Mario Ndreko Marjan Popov, Jose L. Rueda,
Intelligent Electrical Power grids Mart A.M.M. van der Meijden
Faculty of EEMCS Intelligent Electrical Power grids
Delft University of Technology Faculty of EEMCS
Delft, the Netherlands. Delft University of Technology
Delft, the Netherlands.

AbstractThis paper proposes a new knowledge-based control compensation may be required. In such situations, the priority
philosophy for the direct voltage and power control of a multi- is keeping the voltages at all terminals of the MTDC grid
terminal voltage source converter based offshore HVDC grid. strictly within acceptable boundaries established by grid codes
The limitations of the classical direct voltage droop control
strategy are discussed and mainly the difficulty to reach power for which droop is capable [7], [8]. Notwithstanding, droop
reference set-points is stressed. In that context, a knowledge based control strategy while being very exceptional, it has inherent
intelligent controller (namely Fuzzy) is proposed. It is capable drawbacks associated with its implementation. It is known that
of addressing these weaknesses by combining the advantages of droop is a combination of constant voltage and constant power
the droop controller such as robustness and exceptional ability to control [1]. One of the most reported drawback is the voltage
compensate for imbalance during contingencies, and the constant
active power controller which has the ability to easily reach and the power deviations when set points are ordered [1], [5],
power set points. In this context, the power dispatch of the [9].
HVDC grid converters is achieved without the need to solve This problem is not new and several proposals have been
before-hand HVDC grid load flow equations where the droop made in literature [1], [9], [10]. Most proposals have been
constant is included in the algorithm. The advantages of the new analytical in nature, often by means of solving the power flow
Fuzzy controller is the reduced computational effort, the high
degree of flexibility, and the zero percentage error. The efficacy algorithms, whilst including the influence of the droop constant
and robustness of the control strategy is demonstrated by means in these equations. However, these methods that includes the
of time domain simulations for a three terminal voltage source influence of droop constant require computational resource,
converter based offshore HVDC grid system used for the grid they are tedious, and may require a high level of power
connection of large offshore wind power plants. systems expertise. There is also uncertainty with the flexibility
and influence of topology on analytical methods to solve the
Index TermsMTDC grids, voltage control, power control,
gain scheduling, supervisory control, fuzzy control problem.
Alternatively, constant voltage and power whilst very sim-
I. I NTRODUCTION ple, always reach set points without need for pre-calculations
of the MTDC grid load flow. However, such control assumes
Ever since the advancement in power electronic controls that the system always operate under normal conditions [1] at
coupled with the development of Voltage source converters all times. In contingency cases, they cannot compensate for
(VSCs), there have been a vested interest in the concept of power imbalances and will make the MTDC grid unstable.
multi-terminal direct current (MTDC) grids for integration This paper proposes a non-analytical solution to the prob-
of large scale renewable energy sources (RES), bulk power lem of power and voltage deviation in an MTDC grid. The
transfer, and improvement of system security. More important proposed methodology is non-analytical in the sense that,
is the flexibility and controllability of VSCs which could while still employing power flow calculation, it does not
allow for near seamless operation with high voltage alternating include the influence of droop constant in the power flow
current (HVAC) grids with flexibility in operation. There are algorithm. Power flow is only used to obtain direct voltage set
three (3) widely proposed control strategies, viz., constant points that provide the specified power dispatch in the HVDC
voltage, constant power, and droop strategies [1][6]. grid. Fuzzy controller then uses knowledge about the system
Droop control strategy is the most accepted method con- in the form of selected signals that describe the important
sidering its peculiarity with the droop control in high voltage dynamics to transition online from the droop mode to the
AC (HVAC) transmission systems. It is almost a ubiquitous constant power mode (since the set points will be always
term in MTDC literature. It is robust with an exceptional be reached) when new set points are ordered. After steady
response particularly in contingency situations where power state conditions are fulfilled, it transitions autonomously in a
c 2016 IEEE
978-1-5090-3358-4/16/$31.00 smooth manner to droop (as we would have it be in steady
state) and remains in this mode for as long as is required. The Knowledge Base
proposed methodology acts as a gain scheduling system that Inputs Fuzzification
Rule Aggregation Defuzzification
schedules droop gains based on defined conditions such as set
point changes and fault. Fuzzy theory is applied here for its
ability to incorporate highly nonlinear control and combine Fig. 1. Static Map of a Typical Fuzzy Controller
several strategies that is otherwise not possible with a single
analytic equation. Hence, droop constants are highly adaptable
grids where topology or market conditions can change at any
and flexible, without any influence by topology or market
constraints, or size of grid. The proposed solution can be
Fuzzy control is typically applied to feedback control as is
installed as a plug n play box on all droop active terminals.
the case in this paper. The general structure of a knowledge-
This paper is organized as follows, section II gives an
based fuzzy controller is depicted in Fig. 1 and a similar
overview of the state of art of control in MTDC grid, section
structure was applied in this paper.
III give a brief introduction of fuzzy control, section IV
describes the proposed strategy, section V simulation studies,
and finally the conclusions in section VI. DPdc _ droop Active current
component of
idmax the VSC-HVDC
II. C URRENT S TATE OF THE A RT *&) + R - ki _ P
kp_P +
Current proposals are now heading towards the direction - + s
of hierarchical control and coordination from a system level '&) +
*&) ,$ (
perspective. In [11], the authors proposed general control -
strategies for multi-terminal HVDC VSC connected systems '&) ,$ (
and [6], [12], gave a detailed overview of most relevant control
strategies. A priority-based control system was implemented
in [13] which reiterates the influence of the topology, size, and ( + !2)
Fuzzy Supervisor "#$%&
market structure in the implemented control. A drawback of (&'$ ( )2
this priority scheme is the lack of flexibility; for instance, in
the event priority changes from the one converter to another, Fig. 2. Adaptive Fuzzy based droop controller
or in the event where network topology changes.
Similar to the previous, [14] proposed a configuration-
based approach to control of voltage in a 3 terminal DC IV. P ROPOSED C ONTROL S TRATEGY
grid. This is to show the influence of different configurations
A. Strategy and Network
to robustness of control approaches.
Authors in [7] described the problem of distributed slack The objective of this control strategy is to autonomously
bus system where several converters which are installed with combine the capabilities of two well known strategies droop
constant voltage controller may try to act on changes in their and constant power. The classical PI regulator is not changed,
own bus voltages leading to chaotic oscillation of voltage and only the droop gain is accordingly defined by the Fuzzy control
power. The solution proposed by the authors was still to use loop on-line.
a distributed approach but with droop rather than constant The proposed strategy employs a fuzzy controller to transi-
voltage at all distributed terminals [15]. tion from droop control to constant power when new set points
are received to enable system track set points accurately. After
III. M ODEL -BASED F UZZY S UPERVISORY C ONTROL steady state conditions are fulfilled, fuzzy initiates a smooth
L. Zadeh in 1965 put forward the theory behind fuzzy logic transition back to droop. A block diagram of the proposed
and control as we know it today [16]. Fuzzy control is one of control strategy is depicted in Fig. 2. As it can be seen, the
the many computational intelligence methods applied to expert droop gain is scheduled by the output of the fuzzy controller.
control systems. It is also classified under a set of knowledge- The strategy is tested on a three terminal offshore HVDC
based control methods that utilize human knowledge about grid as shown in Fig. 3 presented in [18]. The model consists
a system in order to make control decisions. Fuzzy exploits of two grid side VSC stations (VSCs 2 & 3) connected to AC
the fact that in the real world, physical parameters seldom grids and one VSC station (VSC 1) connected to wind power
possess a definite classification. Fuzzy control is no longer plants. VSC 3 is a slack terminal in constant voltage priority,
a theoretical concept as have been applied to thousands of and VSC 2 in droop. The offshore converter station injects
industrial processes with seemly successful results [17]. It constantly the delivered power by the wind turbines to the
is now a very popular choice for non-linear control, on-line HVDC grid. Fuzzy-based strategy was implemented on VSC
supervisory controller, and adaptive gain scheduler especially 2 in this study. In principle, it can be installed on all droop-
to power system problems. Fuzzy is usually lauded for its active terminals in any grid size. In addition, the proposed
high flexibility and its ability to cope with any level of non- strategy relies only on local measurements such as, terminal
linearity, unanticipated changes, as is the case in HV-MTDC voltage, power signals, and adjacent grid voltage.
Offshore Power Plant
1 1

Membership degree

Membership degree
0.5 OK
1000 MW AC_A 0.5
Df ac 781 MW
Equivalent system
dc 1 Cdc
AC_B 3
Equivalent system Cdc dc
Meq, (3) Meq = 8 0
U dc slack 0
D=3 0 0.005 0.01 0.015 0.02
200 MW 0 0.2 0.4 0.6 0.8 1
2 U (1) Input
dc 325.48kV dc Input

Membership degree
Cdc HVDC cable
U dc(2)
L12=L13=300km ZERO

Fig. 3. Network Model
0 0.1 0.2 0.3 0.5
B. Implementation of Strategy 0.05
Input 0 1
Error GridV
First, a simple power flow is run to determine the set
points that give a specified power flow in the DC grid. Then Fig. 4. Plot of MFs and 3-D Surface Plot
the DC grid is initialized also from power flow. At defined
times, both Udc,ref and Pdc,ref are received simultaneously
by the droop controller. The fuzzy controller then extracts all
the information required to decide whether to transition from The first input is the main signal that combines the error
droop to constant active power, or remain in droop. If condition from both the voltage control block and power control block
satisfies that required for transition to constant power, fuzzy since both errors are sufficient to represent the whole system.
facilitates this. By principle, system change from one set Each of Pe and Ve could be used as separate inputs but
point to another is not instantaneous, and the calculated error combining both dramatically simplifies the strategy. AC grid
reflects this. Hence as error gradually reduces, fuzzy controller voltage is used to detect when there is fault in the connected
(simultaneously as error is reducing) smoothly transitions the AC grid, and derivative of reference power ensures that the
system from constant active power towards droop and when fuzzy controller only transitions to constant power for a
the error is definitely within bounds of steady state condition, reference change. Without this signal to discriminate, the fuzzy
fuzzy completes transition to droop and remains in droop controller will facilitate a transition to constant power for every
except new set points are received. non-zero error detected.
The following procedures were followed to implement the This is important because any change at all in the DC grid,
proposed strategy as is generally used for fuzzy controller including contingency events, will cause a non-zero error at all
implementation. terminals. However, if and only if there is a non-zero error at
1) Experimentation by Simulation: The first step is exper- the measuring terminal and the rate of change of reference is
imentation with model to determine the universe of discourse non-zero, will the fuzzy controller facilitate transition to con-
of input variables. Experimentation also allows for observing stant power, otherwise, system remains in droop. The square
the behaviour of the test system from the time where new set of variables eliminates inherent oscillations, nonlinearity, and
points are ordered, to when set points are reached, and steady dramatically reduces the complexity of implementation
state conditions are fulfilled. reduces number of rules, reduces number of MFs.
2) Inputs and Output: As previously mentioned in section
III the output is the variable to be manipulated. In this case, 3) Membership Functions and Linguistic Terms: MFs were
the droop constant R is the output to be manipulated between selected for each variable. For (Pe2 + Ve2 ) and dPref 2
, the
0-20 (5% droop) as singletons. Determining the input is not sigmoidal MF with two linguistic terms, ZERO and PBIG
very straight forward as initially thought, but a knowledge of were chosen. PBIG implies positive big, and is just a term
the system and the desired control objectives can support in which is highly subjective. Sigmoidal MF was particularly
determining inputs. chosen to imitate a smooth transition from constant power to
These are the inputs for the proposed strategy: droop. Trapezoidal MFs are also suitable but they lead to sharp
edges at transition points which is not acceptable, but there
are ways around this such as normalization. Trapezoidal MF

2 2
(Pe + Ve )

Inputs = Egrid was sufficient for AC grid voltage with two linguistic terms,

2 Faulty and OK (for simplicity). The output is singleton with
dPref gain of 0 or 20 (VLOW or HIGH respectively) at extremes,
Where Pe = Pdc,ref Pdc , Ve = Vdc,ref Vdc , Egrid is but between the transitions, gain vary smoothly as dictated by
the AC grid voltage, and dPref is the rate of change of power the sigmoidal MF of the error signal. The MFs are depicted
reference set point. in the Fig. 4
4) Knowledge-Base and Inference Engine: This is the main -100
part of the implementation. The rules encode the required -200
Proposed Strategy
Droop Only
control actions from the perspective of an operator. The rules Desired Set Point

Power (MW)
usually take the form of, IF...THEN. For this implementation,
the rules have been reduced from 27 possible rules that -400

cover the entire universe of discourse, to 4 rules through -500

mathematical manipulation and elimination of redundant rules. -600

The rules are highlighted below:
IF Error is ZERO AND GridV is OK THEN DroopGain 0 2 4 6 8 10
time (s)
IF Error is PBIG AND GridV is OK AND dPref is ZERO 326
THEN DroopGain is HIGH Proposed Strategy

IF Error is PBIG AND GridV is OK AND dPref is PBIG 324 Droop Only
Desired Set Point

Voltage (kV)
THEN DroopGain is VLOW 322
IF GridV is Faulty THEN DroopGain is HIGH
To keep simplicity, the Mamdani inference was used and
the centre of gravity (COG) method used for defuzzification. 318

Takagi-Sugeno type inference can also be used especially in

the case when the objective is to switch from one mode to 0 2 4 6 8 10

another based on power level. time (s)

Droop Gain
To demonstrate the robustness and efficacy of the proposed
control strategy to the described control objectives, several
Droop Gain

scenarios have been studied viz., set point change, constantly 10

changing set points over a time line, offshore wind side
HVDC station outage, and loss of an onshore HVDC converter. 5
In all scenarios it was assumed that wind power of 1000
MW injected to the HVDC grid. The objective does not 0
0 2 4 6 8 10
include the WF (wind farm) VSC as they usually have a time (s)
more complicated control strategy which was not modelled.
Studies were performed in MATLAB/SIMULINK (R2015b) Fig. 5. System Response to a Set point change at Terminal VSC 2
A. Set Point Change
system back in droop mode. Observe that there is no deviation
To further illustrate the capabilities of the control strategy, at all from ordered set points in steady state.
a set point change scenario was simulated. The proposed
strategy in Fig. 2 was implemented in VSC 2, while VSC B. Constantly Changing Set Points
3 was kept in voltage control priority. Network was initialized To further prove the robustness of the proposed strategy to
as shown in Fig. 3 from a simple power flow calculation. constantly changing set points every 10s over a 50s range,
Now, at 1s new set points were sent to VSC 2 to increase simulation was carried out with set points changing from first
the power from 200 MW (inversion) to 600 MW (inversion). from initial of 200 MW to 400 MW in steps of 100 MW, then
A time constant block was used in the power reference to the reverse is done, and in the final step, set point changed
prevent oscillations and interactions. This in no way affects the from 200 MW (inversion) to 200 MW (rectification). The plot
operation of the proposed strategy. Fig. 5 for VSC 2 shows the shown in Fig. 6 shows how efficient the proposed strategy
comparison between the droop only method and the proposed is in both direction inversion and rectification, without any
strategy and the droop gain adapted by the strategy. As can deviation from ordered set points in steady state. Observe how
be observed, at the moment new set points were ordered, the droop gain is adapted on-line each time new set points are
a definitely large error was detected by the fuzzy controller received.
accompanied with a large rate of change in power reference.
This signified that a change in set points have been ordered. C. Wind Outage
Fuzzy facilitated a change to constant active power (R = 0) and As already described in the course of this paper, the
remained in constant power until both the power and voltage objective is to have droop in steady state in order to ensure
error start to reduce (4.3s) (subplot 3 for droop gains) at which balance is maintained during contingencies to keep voltages
fuzzy controller very smoothly initiated transition to droop. within acceptable limits. In such cases, priority is given to
At about 5s conditions for steady state had been fulfilled and voltage control. So long as new set points are not received, it
400 600 21
VSC 2 (Proposed)
Droop Gain
Proposed Strategy 400 VSC 2 (Droop only)
200 Droop Only 20.5

Power (MW)

Droop Gain
Desired Set Point 200
Power (MW)

0 20
-200 19.5

-400 -400 19
0 2 4 6 8 0 2 4 6 8 10
-600 time (s)
time (s)
0 5 10 15 20 25 30 35 40 45 50
time (s) 1.02 1
VSC 2 (Proposed) VSC 3
VSC 2 (Droop only)
1.01 0.98

Voltage (pu)

Voltage (pu)
Proposed Strategy
1 0.96
Droop Only
330 Desired Set Point
Voltage (kV)

0.99 0.94

0.98 0.92
325 0 2 4 6 8 0 2 4 6 8
time (s) time (s)

320 Fig. 7. System response to total loss of offshore wind power

0 5 10 15 20 25 30 35 40 45 50
time (s)

20 0
Power (MW)
Droop Gain

10 -600

0 1 2 3 4 5 6 7 8 9 10
0 time (s)
0 5 10 15 20 25 30 35 40 45 50
time (s) 21
Droop Gain
Droop Gain

Fig. 6. System response to constantly changing set points


is expected that the system maintain droop mode. As depicted
in Fig. 7, complete loss of wind power was simulated at 1s. 19
0 1 2 3 4 5 6 7 8 9 10
The figure (subplot 2) shows the robustness of the strategy time (s)
in keeping the maximum droop gain (20) as it is required 1.06
to be, thus ensuring that voltages at both VSC 2 and VSC VSC 2
1.04 VSC 3
Voltage (pu)

3 remain within defined bounds (0.9-1.1). Both the proposed

strategy and the conventional droop have exactly matching 1.02

characteristic as it ought to be. 1

D. Loss of Converter 0.98

0 1 2 3 4 5 6 7 8 9 10
In a similar manner as for the loss of wind, loss of time (s)

converter, Fig. 8 shows the efficacy of strategy in keeping

Fig. 8. Complete outage of VSC 2
droop properties so long as there is no set point change. This
is very important in ensuring voltages are maintained strictly
within limits (0.9-1.1) as is depicted from the plots. At 3s,
sudden outage of VSC 2 was simulated. grid power. However, priority is given to voltage control during
contingencies. Subplot 3 shows clearly the voltage barely rose
E. Fault initiated at VSC 2 grid side from the pre-disturbance value for a fault as severe as a bolted
To ensure the proposed strategy is robust for the most fault. Subplot 4 shows the droop gain which remained stable at
contingency scenarios, a fault was simulated by initiating the maximum gain as required to be. As part of the rules base
a 100% dip in voltage for 200ms (1-1.2s) at adjacent grid in Section IV, Rule 4 gave complete priority to droop mode if
connected to VSC 2. This represents complete loss of a power conditions for a fault are detected no matter the previous mode
sink and thus excess power in the DC grid. This leads to the system was pre-fault. This demonstrates the robustness of
voltage rise during the fault. Fig. 9 (subplot 2) shows the AC the strategy.
1000 [6] T. K. Vrana, J. Beerten, R. Belmans, and O. B. Fosso, A classification
1500 I cab1 VSC 2
of DC node voltage control methods for HVDC grids, Electric Power
DC Cable Current (A)

I cab2 800

Systems Research, vol. 103, pp. 137 144, 2013. [Online]. Available:


500 [7] J. Beerten, D. Van Hertem, and R. Belmans, VSC MTDC systems with
a distributed DC voltage control - a power flow approach, in PowerTech,
0 200 2011 IEEE Trondheim, June 2011, pp. 16.
[8] A. Egea-Alvarez, F. Bianchi, A. Junyent-Ferre, G. Gross, and O. Gomis-
-500 0
0.5 1 1.5 2 0.5 1 1.5 2 Bellmunt, Voltage control of multiterminal VSC-HVDC transmission
time (s) time (s)
systems for offshore wind power plants: Design and implementation in
a scaled platform, IEEE Transactions on Industrial Electronics, vol. 60,
1.06 21
no. 6, pp. 23812391, June 2013.
1.04 VSC 2 1
[9] J. Beerten and R. Belmans, Analysis of power sharing and voltage
deviations in droop-controlled DC grids, IEEE Transactions on Power
Voltage (pu)

Droop Gain

1.15 1.2 1.25 1.3

Systems, vol. 28, no. 4, pp. 45884597, Nov 2013.
1 [10] X. Zhao and K. Li, Droop setting design for multi-terminal
19.5 HVDC grids considering voltage deviation impacts, Electric Power
Systems Research, vol. 123, pp. 67 75, 2015. [Online]. Available:
0.96 19
0.5 1 1.5 2 0 2 4 6 8
time (s) time (s)
[11] S. Ruihua, Z. Chao, R. Li, and . X. Zhou, VSCs based HVDC and
its control strategy, in Transmission and Distribution Conference and
Exhibition: Asia and Pacific, 2005 IEEE/PES, 2005, pp. 16.
Fig. 9. Three phase fault at the VSC 2 AC terminals [12] W. Wang, M. Barnes, and O. Marjanovic, Droop control modelling
and analysis of multi-terminal VSC-HVDC for offshore wind farms, in
International Conference on AC and DC Power Transmission (ACDC
VI. C ONCLUSIONS 2012), 10th IET, Dec 2012, pp. 16.
[13] L. Xu, B. Williams, and L. Yao, Multi-terminal DC transmission
This paper studies the efficacy and robustness of a Fuzzy systems for connecting large offshore wind farms, in Power and Energy
based control strategy to the problem of voltage and power Society General Meeting - Conversion and Delivery of Electrical Energy
in the 21st Century, 2008 IEEE, July 2008, pp. 17.
deviation in voltage source converter based MTDC grid. The [14] J. Liang, T. Jing, O. Gomis-Bellmunt, J. Ekanayake, and N. Jenkins,
proposed strategy combines the advantages of the constant Operation and control of multiterminal HVDC transmission for offshore
power controller (in reaching the exact set points required), wind farms, IEEE Transactions on Power Delivery, vol. 26, no. 4, pp.
25962604, Oct 2011.
and all capabilities of droop controller during contingency [15] C. Dierckxsens, K. Srivastava, M. Reza, S. Cole, J. Beerten, and
situations (such as offshore wind power outages and AC R. Belmans, A distributed DC voltage control method for VSC-MTDC
faults). The proposed control strategy in this paper involves systems, Electric Power Systems Research, vol. 82, no. 1, pp. 54 58,
2012. [Online]. Available:
only a simple HVDC grid power flow calculation that do not pii/S0378779611001933
include the influence of the droop constant. [16] L. Zadeh, Fuzzy sets, Information and Control, vol. 8, no. 3, pp. 338
353, 1965. [Online]. Available:
The capabilities of the proposed control strategy is presented article/pii/S001999586590241X
in transitioning between both constant power and droop, and [17] R. Precup and H. Hellendoorn, A survey on industrial applications of
maintaining droop properties so long as set points are not fuzzy control, Computers in Industry, vol. 62, no. 3, pp. 213 226,
2011. [Online]. Available:
changed. For future studies, more attention will be placed pii/S0166361510001363
on proving the stability of the fuzzy controller towards a [18] M. Ndreko, M. Popov, J. L. Rueda-Torres, and M. A. van der Meijden,
actual implementation of the strategy. Proving the stability of Impact of offshore wind and conventional generation outages on the
dynamic performance of AC-DC transmission systems, in PowerTech,
fuzzy control systems is not a new problem and is widely 2015 IEEE Eindhoven. IEEE, 2015, pp. 16.
documented in literature. However, better method of proving
their stability has also been documented.

[1] T. Haileselassie and K. Uhlen, Impact of dc line voltage drops on
power flow of mtdc using droop control, IEEE Transactions on Power
Systems, vol. 27, no. 3, pp. 14411449, Aug 2012.
[2] J. Liang, O. Gomis-Bellmunt, J. Ekanayake, and N. Jenkins, Control
of multi-terminal VSC-HVDC transmission for offshore wind power,
in Power Electronics and Applications, 2009. EPE 09. 13th European
Conference on, Sept 2009, pp. 110.
[3] T. Haileselassie, K. Uhlen, and T. Undeland, Control of multiterminal
HVDC transmission for offshore wind energy, in Nordic wind power
conference, 2009, pp. 1011.
[4] C. Ismunandar, v. d. A. Meer, M. Gibescu, R. Hendriks, and W. Kling,
Control of multi-terminal VSC-HVDC for wind power integration using
the voltage-margin method, in in Proc. 9th International Workshop
on Large Scale Integration of Wind Power into Power Systems as well
as on Transmission networks for offshore Wind Power Plants, Quebec,
Canada, October 2010, pp. 427434.
[5] T. Haileselassie and K. Uhlen, Precise control of power flow in
multiterminal VSC-HVDCs using DC voltage droop control, in Power
and Energy Society General Meeting, 2012 IEEE, July 2012, pp. 19.