978-1-4799-5141-3/14/$31.00 ©2014 IEEE

ANFIS Based Coordinated Control for an Offshore Wind Farm Connected VSC MTDC System *M. Ajay Kumar, Member, IEI, B. Vijay Kumar, Member, IEI, N.V. Srikanth, Member, IEEE

Department of Electrical Engineering National Institute of Technology Warangal, India

*ajaykumar22@nitw.ac.in

Abstract: Offshore wind farms are becoming an attractive solution for wind power in many developed countries. The voltage source converter (VSC) based multi-terminal high voltage direct current (MTDC) transmission system is an attractive technical option to integrate these offshore wind farms with the onshore grid due to its unique performance characteristics and reduced power loss via extruded DC cables. In this paper, an adaptive neuro fuzzy inference system (ANFIS) based coordinated control design has been addressed for MTDC system in order to enhance the reliability and to balance the power. A four terminal VSC-MTDC system which consists of an offshore wind farm and oil platform is implemented in MALAB/SimPower Systems software. The proposed model is tested under different fault scenarios and simulation results show that the novel coordinated control design has great dynamic stabilities and also the VSC-MTDC system can supply Ac voltage of good quality to offshore loads during the disturbances.

Keywords— ANFIS; Coordinated controller; Offshore wind; MTDC; MATLAB; VSC HVDC.

I. INTRODUCTION In recent years, renewable energy sources are becoming

more popular for producing electrical power. Among them, wind energy has become one of the foremost economical and both environmentally and technically attractive options [1]. Wind installations have up to now involved largely on-land sites. However, wind potential seems to exist also in offshore, where there is an advantage of higher wind conditions and fewer environmental restrictions, although the disadvantage of more difficult access and higher installation and maintenance costs must be taken into account [2]. Offshore wind farms are often connected to the main AC grid via high voltage direct current (HVDC) transmission technology. Nowadays, with the growth in HVDC transmission, voltage source converter HVDC (VSC HVDC) also known as “HVDC Light” transmission system has become more and more important in the larger interconnected power system [3]. The main advantage of HVDC Light over conventional HVDC is that the extension to multi-terminal DC (MTDC) systems is relatively easy and hence, the application of MTDC systems is becoming more attractive than before. The VSC based MTDC system is better than the two-terminal HVDC system in several aspects like reliability, control, flexibility and economics [4]-[5]. One essential application of VSC based MTDC transmission system is to interconnect offshore wind farms and oil/gas platforms to the onshore grid, which will reduce the operational costs and increase the reliability. Although, very few number of VSC based MTDC systems installed so far, a large number of

publications exist in the area of VSC based MTDC. All these research works concentrate on different aspects such as DC fault location and protection of MTDC [6], control methodologies [5], [7], and also modeling of MTDC [8].

MTDC systems for power transmission between conventional AC networks and DFIG based wind farms were described in [4], [9]. In [10], a three terminal VSC based HVDC system connecting onshore AC grid to two offshore wind farms was introduced and analyzed. Recently, a four terminal MTDC system was developed [11] where two onshore AC grids located at different geographical area were integrated by two offshore wind farms and the DC grid control strategy and power sharing were clearly depicted. However, as far as control of MTDC system is concerned, the research was constrained to the conventional coordinated control design [12]. In this paper, an intelligent coordinated controller (ANFIS based) is implemented for the first time in MTDC systems, which gives fast response and good quality of supply to offshore platforms without any mathematical modeling.

II. MTDC SYSTEM MODEL The multi-terminal VSC based HVDC system tested in this

paper is shown in Fig.1. It consists of four terminals, one terminal (VSC1) is connected to the conventional power plant feeding power via VSC2 to a strong AC grid located at city centers through HVDC Light transmission system, the third terminal (VSC3) is connected to an offshore DFIG based wind farm transmitting power to the onshore grid via DC cables and fourth converter (VSC4) links to the offshore platforms supplying power from the dc link.

AC Filters

AC Filters

AC Filters

AC Filters

DC

Cab

le(5

0km

)

Fig.1 MTDC system for offshore wind farm and oil platform interconnection

The main objectives of controlling MTDC system is not only to improve the overall performance of the system, but also to protect the equipment which is in service. VSCs play a vital role in the safe operation of the MTDC system.

VSC1 controls the active power and reactive power where as VSC2 adopts a DC voltage control method. But the wind farm side converter VSC3 must use constant active power and AC voltage, using power independent control systems. Finally, the converter connected to an oil platform (VSC4) is adopted with AC voltage control method in order to provide uninterrupted and balanced AC voltage at the terminal. Each VSC of the four terminal MTDC system is coupled with AC network via line resistor R, phase reactor L and a DC capacitor C is in parallel to the DC bus from the station as shown in Fig.1. The following equations are obtained in the d-q synchronous frame [13].

Fig.2. Control structure of a converter

qd

dcdsd Li Ri

dtdiLVV ω++=− (1)

dq

qcqsq LiRi

dtdi

LVV ω−+=− (2)

Where Vsd and Vsq are source voltages, id and iq are line currents, Vcd and Vcq are converter input voltages. Based on the instantaneous power theory, neglecting the losses of the converter and the transformer, the active and reactive power exchanges from the AC end of the DC link are:

)iViV(23P qsqdsdac += (3)

)iV-iV(23Q dsqqsdac = (4)

Suppose, the direction of the source voltage vector as d-axis,

Vsq =0. So (3) and (4) can be re-written as:

dsdac iV

23P = (5)

qsdac iV

23Q = (6)

Since Vsd is constant, from (5) and (6) it is clear that the active power will be controlled by id , whole the reactive power will be controlled by iq. On the DC side of the converter, DC current and DC power are;

c

dcdc i

dtdvCi += (7)

dcdcdc iVP = (8) Where idc is the DC current to be followed by the capacitor,

vdc is the DC link voltage and ic is the current on the DC cable. Neglecting the loss of converter, power of AC side equals to the DC side.

dcac PP = (9)

dcdcdsd iViV

23 = (10)

Based on the law of conservation of energy, the active power transferred in the MTDC system must satisfy the following equation:

P1+P2+P3+P4 = 0 (11)

In MTDC system, each VSC is controlled by local controller and the whole system is coordinated by the master controller. Now, the control methodologies of MTDC system are briefly discussed as follows:

A. Outer Controllers In general, constant active, reactive power control or

constant AC/DC voltage control strategy can be adopted for local control at each VSC in MTDC system. Control circuit of VSC is shown in Fig.2 consists of an outer control loop and inner current control loop. The outer controller includes the active, reactive power control, DC voltage control and an AC voltage controller. The choice among these controllers will depend on the application. The outer controller will calculate the reference values of the converter current.

From the equations (5) and (6), it is clear that every conver-ter can control its active and reactive power independently. A combination of an open loop and PI controller is used to keep the active power to its desired value, given by the equation:

(12) P)P)(

sKK(

vPi ref

ip

sd

refsd_ref −++=

Similarly, reactive power can also be controlled as in the

previous case by combining the PI controllers as shown in below.

(13) Q)Q)(s

KK(vQ

i refi

psq

ref_ref sq −++=

In general, the AC voltage controller is chosen at inverter station located on offshore oil platform so as to obtain an uninterrupted and balanced AC voltage from the AC voltage controller, the d-axis current reference can be obtained using the equation

(14) )vv)(

sKK(i ss_ref

ipd_ref −+=

With v)v(vv sd2si

2sds =+=

(15) )vv)(

sKK(i sds_ref

ipd_ref −+=

MTDC system should maintain a constant DC link voltage

under normal condition in order to satisfy the power balance equation. When the MTDC system active power is super flows, VSC2 sends back to the AC grid and in this way without any

energy storage device, VSC2 acts as an energy buffer by encountering the switching losses and transmission losses. When a PI controller is used, the DC current reference of VSC2 can be written as

(16) )vv)(

s(i dcdc_ref

ipdc_ref

KK −+=

All these outer loop PI regulators calculate the reference

value of the converter current vector (Iref_dq), which is the input to the inner current loop.

B. Inner Current Controller According to the equations (1) and (2), the currents of d and

q axis can be controlled by Vcd and Vcq respectively. The Inner current loop block contains two PI regulators that will calculate the reference value of the converter voltage vector (Vref-dq). By using clarke’s transformation Vref-dq is transformed into Vref-abc, which is the input to the space vector pulse width modulation block.

C. ANFIS based Coordinated Controller MTDC Structure is more complex due to the intercon-

nection of more than two converters for the same DC bus. So it is necessary to control the DC link voltage within acceptable limits to assure that all active power on the DC grid is transmitted into the AC grid/load. In order to ensure the stability and reliability of the MTDC system, ANFIS based coordinated control strategy is introduced in this paper.

ANFIS is an adaptive network that is functionally equivalent to a fuzzy inference system, where the output has been obtained by using fuzzy rules on inputs. Fig. 3 depicts a two - input one - output ANFIS structure [14]. The two inputs are x1 (error) which was obtained as (Vdc_ref ~Vdc), x2 (change of error) and the output is a controlled DC link voltage. Each input and output variable has five linguistic variables, i.e Negative large (NL), Negative small (NS), Zero (Z), Positive small (PS) and Positive large (PL). Hence, in this proposed ANFIS controller total 25 linguistic variables for the output are employed.

Fig. 3. A Five layer ANFIS structure

The proposed ANFIS architecture consists of five layers wherein circle shaped nodes are called fixed nodes, which means the node parameters are independent on the other nodes and square shaped nodes are called adaptive nodes, whose node parameters depend on the other nodes. Each neuron in the first layer corresponds to a linguistic variable while the output

equals the membership function of this linguistic variable. In the second layer, each node multiplies the incoming signals and sends out the product that represents the firing strength of a rule. Each node in the third layer estimates the ratio of the rule firing strength to sum of the firing strength of all rules. In the fourth layer, the output is the product of the previously found relative firing strength of the i th rule. The final layer computes the overall output as the summation of the incoming signals.

The proposed controller is checked in MATLAB/ANFIS editor tool box with a triangular membership function as it offers minimum training error. Since, the back propagation algorithm is notorious for its slowness and tendency to become trapped in local minima, a hybrid learning algorithm is used in this contribution. This algorithm is fast and accurate in identi-fying the parameters. The parameters of the ANFIS controller are given in Table 1.

TABLE.1. ANFIS PARAMETERS Number of nodes 75 Number of linear parameters 75 Number of nonlinear parameters 30 Total number of parameters 105 Number of training data pairs 600 Number of testing data pairs 100 Number of fuzzy rules 25

III. SIMULATION AND RESULT ANALYSIS In the test system as shown in Fig.1, offshore wind farm

consists of 133 units of DFIG; each one has a nominal power rating of 1.5 MW, therefore the total capacity of 200 MW. The onshore AC grid is modeled as 2000 MVA, 220 kV voltage source. The offshore platform is modeled as a passive load of 100 MW. The Test system was implemented in MATLAB/ Simulink and three case studies were carried out to demonstrate the feasibility of the controllers such as a three phase to ground fault on the grid side terminal, change in wind speed, and a three phase fault at the offshore platform. Since, the fault on the ac side of VSC1 does not produce a destructive voltage surge and pose a great threat to the system stability; only the simulation results associated with an ac fault on the inverter side are presented. The dynamic responses of active power, reactive power, voltage and current at all the terminals for different perturbations are shown in Figs. 4-6.

A. Three Phase fault at an AC grid (Near VSC2) A balanced three phase fault was applied on ac grid side at

4s for 5 cycle duration to observe the effect of the proposed controller and simulation results were shown in Fig. 4. It is clear from the Fig. 4(a) to (d) that, active power and reactive powers were quickly able to track their pre-defined values after removing the fault at an AC grid and during the fault there is a possibility of power transfer at all the VSCs expect VSC2. Irrespective of the fault, the offshore wind farm is generating its rated power of 200MW at a constant wind speed of 12m/s as depicted in Fig. 4(e).The DC link voltage of MTDC system and power available at the DC side of the VSCs are shown in Fig. 4(f). Here, one can observe that the DC power transferred through VSC2 during the fault is zero and after the fault is cleared at 4.1s, the VSC2 moves out of the current limit control mode and the DC link voltage oscillations were reduced in 10ms.

2 3 4 5 6 7 80

1

Active power, Pref (pu)

2 3 4 5 6 7 8-0.5

00.5 Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)

PPref

QQref

(a)

2 3 4 5 6 7 8-1

0Active power (pu)

2 3 4 5 6 7 8-101

Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8

-101

Time (s)

Iabc (pu)

QQref

(b)

2 3 4 5 6 7 80

1Active power, Pref (pu)

2 3 4 5 6 7 8

-0.20

Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)

PPref

QQref

(c)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)2 3 4 5 6 7 8

-101

Vabc (pu)

2 3 4 5 6 7 8

-1-0.5

Active power (pu)

2 3 4 5 6 7 8

-0.10

0.1Reactive power, Qref (pu)

QQref

(d)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Iabc (pu)

2 3 4 5 6 7 80

100200

P (MW)

2 3 4 5 6 7 8-50

050

Q (Mvar)

2 3 4 5 6 7 8110011501200

Vdc (V)

2 3 4 5 6 7 8111213

Time (s)

Wind Speed (m/s)

(e)

2 3 4 5 6 7 8

-1

0

1x 10

5 Vdc (V)

2 3 4 5 6 7 8

-1

-0.5

0

0.5

1

Time (s)

DC Power (pu)

VSC1VSC2VSC3VSC4

+Vdc-Vdc

(f)

Fig. 4 Dynamics of (a) VSC1 (b) VSC2 (c) VSC3 (d) VSC4 (e) Wind farm and (f) DC link parameters of MTDC system for a 3-phase fault at onshore ac grid with a duration of 5 cycles.

B. Change in wind speed The second case study is based on the operation with

change in the wind speed at that offshore power plant, and the results are shown in Fig.5. Initially, the wind speed is 12m/s with generated power being around 200 MW from the DFIG wind farm. When the wind speed has been decreased gradually from 12 m/s to 10 m/s and again set to 12 m/s as shown in Fig.5(f), the wind power generated is also changed from 200 MW to 140 MW and accordingly. As the VSC3 can balance ±100 MW fluctuations, irrespective of change in wind speed the active power and reactive powers at all the terminals are smooth and able to track their reference values as shown in Fig.5(a) to (d) and DC power balance can be observed from Fig. 5(e) respectively.

2 3 4 5 6 7 80.9

1

Active power, Pref (pu)

2 3 4 5 6 7 8-0.1

00.1

Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time(s)

Iabc (pu)

PPref

QQref

(a)

2 3 4 5 6 7 8-1

0Active power (pu)

2 3 4 5 6 7 8-0.2

-0.1

0Reactive power, Qref (pu)

2 3 4 5 6 7 8-1

0

1Vabc (pu)

2 3 4 5 6 7 8-101

Time(s)

Iabc (pu)

QQref

(b)

2 3 4 5 6 7 80.7

0.8

Active power, Pref (pu)

2 3 4 5 6 7 8

-0.2

0Reactive power, Qref (pu)

2 3 4 5 6 7 8-1

0

1Vabc (pu)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)

QQref

PPref

(c)

2 3 4 5 6 7 8-1

-0.5

0Active power (pu)

2 3 4 5 6 7 8-0.1

0Reactive power, Qref (pu)

2 3 4 5 6 7 8-1

0

1Vabc (pu)

2 3 4 5 6 7 8-1

0

1

Time(s)

Iabc (pu)

QQref

(d)

2 3 4 5 6 7 8

-1

-0.5

0

0.5

1

Time(s)

Pdc(pu)

VSC1VSC2VSC4VSC3

(e)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Iabc (pu)

2 3 4 5 6 7 8100200 P (MW)

2 3 4 5 6 7 8-50

050

Q (Mvar)

2 3 4 5 6 7 8110011501200

Vdc (V)

2 3 4 5 6 7 810

12

Time(s)

Wind Speed(m/s)

(f)

Fig. 5. Dynamics of (a) VSC1 (b) VSC2 (c) VSC3 (d) VSC4 (e) DC powers and (f) Wind farm parameters of MTDC system for a change in wind speed.

2 3 4 5 6 7 80.5

11.5

Active power, Pref (pu)

2 3 4 5 6 7 8-0.1

00.1

Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)

PPref

QQref

(a)

2 3 4 5 6 7 8-1.2

-1-0.8

Active power (pu)

2 3 4 5 6 7 8-0.2-0.1

0Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time(s)

Iabc (pu)

QQref

(b)

2 3 4 5 6 7 80.60.8

1Active power, Pref (pu)

2 3 4 5 6 7 8-0.3

-0.2Reactive power, Qref (pu)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Time(s)

Iabc (pu)

PPref

QQref

(c)

2 3 4 5 6 7 8-101

Time (s)

Iabc (pu)2 3 4 5 6 7 8

-101

Vabc (pu)

2 3 4 5 6 7 8-1

-0.50

Active power (pu)

2 3 4 5 6 7 8

-0.20

Reactive power, Qref (pu)

QQref

(d)

2 3 4 5 6 7 8-1

0

1x 10

5 Vdc (V)

2 3 4 5 6 7 8-1

0

1

Time(s)

DC Power (pu)

VSC1VSC2VSC4VSC3

+ Vdc- Vdc

(e)

0 1 2 3 4 5 6 7 80

0.5

1

Time(s)

Pitch_deg

(f)

2 3 4 5 6 7 8-101

Vabc (pu)

2 3 4 5 6 7 8-101

Iabc (pu)

2 3 4 5 6 7 80

100200

P (MW)

2 3 4 5 6 7 8-40-20

0Q (Mvar)

2 3 4 5 6 7 8110011501200

Time(s)

Vdc (V)

(g)

Fig. 6 Dynamics of (a) VSC1 (b) VSC2 (c) VSC3 (d) VSC4 (e) DC link (f) pitch angle and (g) Wind farm parameters of MTDC system for a 3-phase fault at offshore oil platform with a duration of 5 cycles.

C. Three Phase fault at an Offshore Platform (Near VSC4) The last case observed was a three phase short circuit fault

at the offshore platform for the duration of 5 cycles, and the simulation results were shown in Fig. 6. Before fault, power is transferred from conventional power plant and offshore wind farms to the AC grid and offshore platform smoothly and during the fault, the power transfer to offshore load is zero and at the remaining terminals power is shared in such a way to get the power balance as shown in Fig. 6(a) to (d). It is clearly demonstrated in Fig. 6(e) that, the DC link voltage of MTDC system has minimum oscillations when compared to a fault at AC grid and after clearing the fault, VSC2 controls the DC voltage and other converters controls the active power hence, the system will reach its normal state immediately. The wind farm dynamics can be observed in Fig. 6(f) and (g) where the pitch angle, wind power and DC voltage of the wind farm converter are stable irrespective of the fault at offshore platform. Even under such large perturbations, the MTDC system continues to operate smoothly in accordance with the pre-defined standards.

IV. CONCLUSIONS In this paper, ANFIS based coordinated control strategy has

been attemted for MTDC system dealing with offshore VSCs and onshore VSCs at different perturbations. The simulation results under normal condition shows that the coordinated control strategy keeps AC voltage RMS constant, transmits the power from wind farms to AC grid and offshore oil platform via DC line. Similarly, for large perturbations at an AC grid/ offshore platform, the MTDC system responds quickly and then each controlled output returns to its pre-defined value immediately. Another advantage of MTDC system operation in integrating offshore wind farms to the onshore grid is that, when parts of the system are separated due to various reasons, the system continues to operate in stable region.

REFERENCES [1] Leon Freris, and David Infield, Reneawable energy in power systems,

Ist ed.. John Wiley & Sons, Ltd, UK, 2008. [2] P. Bresesti, W. L. Kling, R. L. Hendriks, and R. Vailati, “HVDC

connection of offshore wind farms to the transmission system,” IEEE Trans. Energy Conv., vol. 22, no. 1, pp. 37–43, Mar. 2007.

[3] M. Ajay kumar, K. U. Archana, and N. V. Srikanth, “HVDC Light Systems: An overview,” International Review on Modelling and Simulations, Vol. 5, No. 5, pp.1951-1959, October 2012.

[4] W. Lu and B.T. Ooi, “Premium quality power park based on multiterminal hvdc,” IEEE Transactions on Power Delivery, vol. 20, no. 2, pp. 978 – 983, April 2005.

[5] T. D.Vrionis, X.I.Koutiva, and G.B.Giannakopoulos, “Control of an HVDC link connecting a wind farm to the grid for fault ride-through enhancements”, IEEE Transactions on power systems, Vol 22, no.4, pp. 2039 – 2047, November 2007.

[6] L. Tang and B.T. Ooi, “Protection of VSC-multi-terminal HVDC against DC faults,” in IEEE Annual Power Electronics Specialists Conference, vol. 2, pp. 719–724, 2002.

[7] C. Hairong, W. Chao, Z. Fan, and P. Wulue, “Control strategy research of VSC based multiterminal HVDC system,” in IEEE PES Power Systems Conference and Exposition, pp. 1986-1990, 2006.

[8] S. Cole , J. Beerten and R. Belmans "Generalized dynamic VSC MTDC model for power system stability studies", IEEE Trans. Power Syst., vol. 25, no. 3, pp.1655 -1662, 2010.

[9] M. Ajay kumar, K. U. Archana, and N. V. Srikanth, “Fast Fault Recovery of a Grid Integrated Wind Farm Based HVDC Light Transmission System,” in. proc. IEEE/PES Innovative smart grid technologies conference, Kuala Lampur, pp.1-6, 2014.

[10] Ling Xu,Lingling Fan,and Zhixin Miao, “Modeling and Simulation of Multi-Terminal HVDC for Wind Power Delivery,” in proc. 2012 IEEE Power Electronics and Machines in Wind Applications, pp.1-6, 2012.

[11] X. Lie , Y. Liangzhong and M. Bazargan "DC grid management of a multi-terminal HVDCtransmission system for large offshore wind farms", in Proc.Int. Conf. Sustainable Power Generation and Supply, pp.1 -7 2009

[12] C. Dierckxsensa, et al, “A distributed DC voltage control method for VSC MTDC systems,” Electric Power Systems Research, Vol. 82, pp. 54– 58, 2012.

[13] M. Ajay kumar, and N. V. Srikanth, “Performance of SVPWM based vector controlled HVDC Light transmission system under balanced fault condition”, in proc. of IEEE power and energy conference at Illinois, USA, (Feb. 22-23, 2013), 2013, pp. 194-199.

[14] M. Ajay kumar, and N. V. Srikanth, “An adaptive neuro fuzzy inference system controlled space vector pulse width modulation based HVDC light transmission system under AC fault conditions”, Central European Journal of Engineering, Springer, Vol.4. No.1, PP.27-38, March. 2014.