DC Distribution System Architecture and Controls for Wind Power Applications

Yogesh Patel

Rockwell Automation Mequon, USA

yppatel@ra.rockwell.com

Adel Nasiri

University of Wisconsin-Milwaukee Milwaukee, USA

nasiri@uwm.edu; nasiri@uwm.edu

Abstract- In this paper, Medium Voltage DC (MVDC) distribution system is proposed for wind power applications. The proposed MVDC system is favorable for off shore wind farms. The architecture of the system, control for the medium voltage three phase inverter, active and reactive power control for voltage and frequency support are discussed in the paper. The proposed topology allows for easier integration of energy storage with wind energy. The simulation results show the viability of the proposed system for active/reactive power controls for wind farms.

NOMENCLETURE Lg Grid side inductor in LCL filter Lc Converter side inductor in LCL filter L Total inductance (Lg+Lc) Cf LCL filter capacitor Rf LCL filter capacitor series resistor Kp, Ki PI regulator gains Ts Sampling time Ti Time constant of PI regulator Kpwm Gain of PWM module I(s) Single phase equivalent current Van(s) Single phase equivalent voltage Vdc DC link voltage Vd, Vq (d-q) axis voltage at grid side Id, Iq (d-q) axis current at grid side Idref, Iqref Reference (d-q) axis current ω Angular frequency (rad/sec) f Frequency (Hz) p,q Active and reactive power pref, qref Reference active and reactive power Pmin, Pmax Minimum & Maximum active power Qmin,Qmax Minimum & Maximum reactive power

I. INTRODUCTION

The electric energy generation using wind turbines has mostly been in form of AC. This is because most of the wind farm installations are onshore. In addition conventional machines, power conversion systems, control systems and protection schemes have been adapted for wind energy systems. The offshore wind farm installations have been continuing to grow rapidly. Some leading countries in the wind energy area are focusing more on offshore technology, mainly due to lack of the suitable onshore sites, getting closer to loads and

better wind conditions. Large offshore wind farms may be installed 50-60 miles away from the grid connection point onshore. AC distribution through submarine cables is costly due to issues related with high charging currents, reactive power and harmonics. DC power distribution systems can offer several benefits over AC systems including lower losses and voltage drops. Cost effective control and protection of DC distribution system are discussed in the literature [1]. For offshore wind power application, MVDC system is more suitable from size, cost and efficiency point of view over high voltage DC system (HVDC) [2].

With advancement in silicon technology, higher voltage self-commutated switching devices are available in market. This enables supporting higher power medium voltage converters for MVDC system. The medium voltage converters are classified into current-source and voltage-source topologies, depending on the DC-link energy storage component. Voltage source inverters (VSI) are increasingly popular for MVDC application. The VSI can be further classified in multiple topologies like multi-level converters, flying capacitors and cascade H-bridge [3]-[5]. VSI operates at higher switching frequencies generally in range of 2 kHz – 4 kHz to reduce the input LCL filter size.

II. PROPOSED MVDC DISTRIBUTION SYSTEM Figure 1 shows the novel architecture of the DC

distribution system studied in this paper. The power generated by the generators is converted to DC form by SCR-based rectifiers to lower the cost. The power is collected from all the turbines and transferred to the grid side converters. Multiple large size inverters transfer the power to the utility grid. The generator side converters adjust the power from the turbines and can place them on Maximum Power Point Tracking (MPPT). The grid side inverters adjust the DC voltage of the system and perform grid side functions such as exporting reactive power. Energy storage components (ultra-capacitor and batteries) are directly connected to the DC system. They exchange power with the DC system to perform power ramp rate control, power smoothing, power shifting, and transient stability control at the farm output. The typical capacity factor of a wind farm is around 30-35%. The additional capacity of the inverter can be used to operate the energy storage elements when the system is not generating nominal power.

978-1-4673-0803-8/12/$31.00 ©2012 IEEE 3493

Figure 1. The architecture of the proposed DC distribution wind farm.

ssT+11

i

ip sT

sTK +1

21 s

pwm

Ts

K

+ )]()([1

2

2

cgcgfffcg

fffg

LLLLRSCCLLSSRSCCLS

++++++

Figure 2. Simplified single phase equivalent block diagram of VSI with LCL filter current control

III. SELECTION OF THE INVERTER AND FILTER TOPOLOGIES

A three-phase 3-level Neutral Point Clamp (NPC) inverter is chosen as inverter of choice for this application. The 3-level NPC provides higher quality output voltage and current waveform and requires a reduced output LCL filter size and cost compared with two level inverters. Only half of the DC bus voltage has to be switched, which leads to reduced switching losses and higher efficiency. It also creates less Common Mode (CM) current at the output to grid.

Some Space Vector Modulation (SVM) techniques can be utilized to virtually reduce the CM noise to zero. Control loops are also easier to implement. Availability of IGBT modules with ratings of 3.3kV, 4.5kV and 6.6kV makes 3-level NPC topology very popular for MVDC (2.3kV-4.6kV) applications [3]. For this application, the switching frequency is around 2-4 kHz. An LCL filter is designed for the inverter so the output current profile meets the IEEE 1547 standard [6]-[7]. A passive damping resistor is calculated to prevent the resonance in the filter [8].

IV. CONTROL TECHNIQUE FOR MVDC SYSTEM

Two techniques are employed to control the inverter. 1. Constant DC bus mode of operation 2. Direct active/reactive power control. In normal operation, when energy storage is not connected to the DC bus, the active power control is used to regulate the DC bus voltage. The current control forms the inner loop where is the voltage control will form the outer loop. The current control loop equations are given as below [9]-[10]:

dqddrefi

pd eLiiiSKKv +++⎟

⎠⎞

⎜⎝⎛ +−= ω)(

(1)

dqqrefi

pq LiiiSKKv ω−+⎟

⎠⎞

⎜⎝⎛ +−= )(

(2) In this case, the DC bus voltage is controlled using the VSI side DC current. For the voltage control loop the equation (3) can be reached.

GeneratorSide

Converter

GeneratorSide

Converter

GeneratorSide

Converter

UtilityGridBattery Charger/

Discharger

Battery EnergyStorage

UltraCapacitors

Grid SideConverter

Grid SideConverter

Active/ReactivePower Control

Active/ReactivePower Control

3494

( )dcrefdci

pLdcdref VVSKKiii −⎟

⎠⎞

⎜⎝⎛ +=−≈ (3)

For the control loop analysis, its assume that equations (1) and (2) are decouple and consider single phase equivalent control loop of the VSI with LCL filter is shown in Figure 2. In this figure, the transfer function of current loop PI regulator, PWM module and filter with passive damping are indicated [9]-[11].

If energy storage is utilized on the DC bus, the active power output can be regulated independently, considering the state of charge of the storage and the power coming from the wind farm. In this case, several techniques such as power smoothing, power ramp rate control, grid frequency support, and/or grid reactive power support can be applied. Frequency support and grid voltage support is discussed in this paper as follows.

A. Frequency droop and voltage droop control The proposed MVDC system can provide frequency

and voltage support to the grid. Typically, renewable energy systems ride on grid frequency and voltage and do not provide significant ancillary services. However, as renewable installed capacity is rising to become a significant part of the total grid capacity, they must participate in grid support functions. There are several mechanisms including generator governor, automatic gain control and load shedding. The proposed system can provide an alternative mechanism, which can be applied faster than conventional methods. The proposed MVDC system with ultra-capacitors and battery backup with DC/DC converter can allow the VSI to support frequency droop for short duration before automation generation control take appropriate action. Similarly it can support the system to keep the AC voltage within range by regulating the reactive power. Variation in real power lead to change in frequency and variation in reactive power leads to variation in the voltage magnitude of the nodes shown by equation as below [12].

( ) ( )21

2121 2 VV

XPPtffπ−=−

(4)

( )1

21 VXQVV =−

(5)

Frequency susceptibility factor Kpf and voltage

susceptibility factors Kqv can be defined based on

equations (4) and (5) as below.

fPK pf Δ

Δ= , VQKqv Δ

Δ= (6)

The grid supply is mostly dominated by the synchronous generator. Synchronous generator inertial constant is given by equation (7).

MVA

JH

2

21 ω

= (7)

Where H = Inertia constant in MWs / MVA, J is

moment of inertia in kgm2, ω is nominal speed of rotation in rad/s and MVA is the rating of the machine. The inertial response of a synchronous generator is the initial power injection to the system following a change in system frequency caused by a disturbance such as a loss of generation or sudden increase in demand. The Inertia Constant is defined as the stored energy in the rotating mass at the rated speed in MWs/MVA. The majority of generators connected to grid have an inertia constant H of between 3MWs/MVA and 9MWs/MVA. After a system disturbance which results in an imbalance between supply and demand, the inertia prevents an instantaneous change in speed. The rate of change of the speed will be governed by the equation (8):

HPf

2Δ=Δ

(8)

PmaxPmin Prated

fmin

frated

fmax

f

f

P P

Figure 3. Frequency vs. active power and voltage vs. reactive power

characteristics.

Frequency susceptibility factor KPf can be define in term of inertia constant as below:

3495

HKpf 2= (9)

Figure 3 shows the frequency verses active power and voltage verses reactive power characteristic. Slope of the characteristic will be defined based on the grid regulation requirements and limitations, which is nothing but inverse of the frequency susceptibility factor and voltage susceptibility factor with min and max limit constraint.

The single phase equivalent block diagram of frequency droop control and voltage droop control are shown in figure 5 and figure 6. The block diagram is very similar to the current control block diagram as show in figure 2. For frequency droop control, the difference between reference frequency (fref) and measured frequency (f(s)) is converted to Pref using a PI regulator. The gain of the PI regulator is defined by frequency susceptibility factor Kpf. The reference current is calculated from this reference power and actual power measure. For voltage droop control, difference between reference voltage (Vref) and measure voltage (Vs) is converted to Qref using PI regulator. The gain of the PI regulator is nothing but the voltage susceptibility factor Kqv. The reference current is calculated from reference reactive power and measured reactive power. These single phase equivalent block diagram is used to tune the control loop.

Open loop stability analysis performed using single phase equivalent frequency droop control. Bode plot of the frequency droop control loop shown in figure 4. Analysis shows that available gain margin is 25dB and phase margin is 135dB. Similar analysis performed for the voltage support control. Actual frequency droop control and voltage droop control are implemented as shown in Figure 7. Error in frequency will be amplified by the Kpf and using PI regulator provides additional

reference power required to meet the grid reference frequency. Similarly the error in voltage will provide in term of the reactive power required to meet the grid conditions. Active and reactive power control forms an inner loop. Equations for the inner loop are given as below.

drefi

pd eLqppSKKv +−+⎟

⎠⎞

⎜⎝⎛ +−= ω)(

(10)

LpqqSKKv ref

ipq ω−+⎟

⎠⎞

⎜⎝⎛ += )(

(11)

Figure 4. Bode plot of frequency droop control loop.

ssT+11

i

ip sT

sTK +1

21 s

pwm

Ts

K

+

)]()([1

2

2

cgcgfffcg

fffg

LLLLRSCCLLSSRSCCLS

++++++

1

11i

ipf sT

sTK +dd IVP

23=

dd IVP23=

Figure 5. Single phase equivalent block diagram of the frequency droop control.

3496

ssT+11

i

ip sT

sTK +1

21 s

pwm

Ts

K

+

)]()([1

2

2

cgcgfffcg

fffg

LLLLRSCCLLSSRSCCLS

++++++

1

11

i

iqv sT

sTK +qd IVQ

23=

qd IVQ23=

Figure 6. Single phase equivalent block diagram of the voltage droop control

θ

Lωθ

θ

Lω

Figure 7. Control diagram of VSI with Frequency and Voltage droop

support.

V. SIMULATION RESULTS The inverter with LCL filter and control techniques

are implemented using Matlab Simulink. The rated power of the inverter is 600kW and the results are displayed in per unit. Simulating frequency droop control requires variable frequency source, which are dependent on demand and supply of the active power. The variable frequency and voltage source was developed in Matlab Simulink to simulate the system. The block diagram of the simulated system is shown in figure 8. Difference between Pgen and Pload translate in to equivalent frequency variation and difference between Qgen and Qload result in to voltage amplitude variation. The variable frequency and voltage source connected to VSI using LCL. VSI is fed by the DC/DC converter which is supply by the battery backup system. DC/DC converter keeps the DC bus voltage constant. For simulation purpose, wind turbine and converter system ignored. VSI

control includes frequency and voltage droop control. If the frequency or voltage of the variable source varies based on the active- reactive power demand and supply variation, the control will inject or extract active and/or reactive power from source to keep the frequency and voltage to reference value. For simulation purpose it is assumed that the DC source is infinite but in actual application it will be limited and support the frequency and voltage droop for short duration until the automation generation control will respond.

Figure 8. Block diagram of the simulated system in Matlab Simulink

The simulation results for frequency droop control are shown in Figure 9. The waveforms, from top, are phase voltage, line current, and actual power generation Pg Vs actual power demand PL, and converter side line to line voltage. The last waveform contains reference frequency, actual frequency with control and without control. One

3497

can observe that when the active power demand PL increase, the frequency droop control generate active power from DC storage system and inject to the source to meet the demand, which allow frequency closely match with the reference frequency. In last waveform of the frequency control simulation result shows that with frequency droop control the actual frequency (59.9Hz) is very closely follow the reference frequency (60Hz). In absence of the frequency droop control, as demand increase and not additional active power available, the frequency drops down (59.6Hz) as shown in figure9.

Voltage droop control simulation results are shown in figure 10. The first and second waveforms are the phase voltages with and without voltage support control. The voltage sag with voltage support control is less than 10% whereas without is about 33%. The third waveform shows reference and actual reactive power for the control loop. The generated and total demand of the reactive power is shown in fourth waveform. It can be easily observed that when the reactive power demand increases, the reference for the reactive power control goes high and generate more reactive power to keep the phase voltage constant. In absence of the control the phase voltage sag is higher.

VI. CONCLUSIONS In this paper, a proposed MVDC system is discussed

in detail. The main advantages of the proposed MVDC system are active and reactive power rate control, frequency droop control and voltage droop control. The selection of the basic building blocks for MVDC system such as VSI / VSC and filter are considered in this paper. For the different control approach like constant dc bus control, active reactive power control and frequency-voltage droop support control, mathematical model of the control loops are developed. Single phase equivalent model are analyzed using Matlab Simulink control design environment. The system simulation result is closely matched with the analysis. The simulation results also verify that wind farms with MVDC system including control of the frequency and voltage droop support help the grid system.

VII. ACKNOWLEDGMENT

The authors acknowledge the financial support provided for this project under contract MIL104452 by Wisconsin Energy Research Consortium (WERC).

Figure 9. Simulation results for frequency droop control.

3498

Figure 10. Simulation results for voltage droop control.

REFERENCES [1] M. S. Carmeli, F. Castelli-Dezza, D. Rosati, G. Marchegiani,

M. Mauri, “ MVDC connection of offshore wind farms to transmission system,” IEEE power electronics electrical drives automation and motion, pp 1201-1206, 2010

[2] Wulue Pan, Yong Chang, Hairong Chen, “Hybrid multi terminal HVDC system for large scale wind power,” IEEE power systems conference and exposition, pp 755-759, 2006

[3] D. Kurg, S. Bernet, S. Fazel, K. Jalili, M. Malinowski, “Comparison of 2.3kV medium voltage multilevel converters for industrial medium voltage drives,” IEEE Trans. Industrial electronics., vol. 54, no. 6, pp. 2979–2992, Dec. 2007.

[4] J. Rodriguez, J. S. Lai, and F. Z. Peng, “Multilevel inverter: A survey of topologies, control, and applications,” IEEE Trans. Ind. Electron., vol. 49, no. 4, pp. 724–738, Aug. 2002.

[5] J. Rodriguez, S.Bernet, Bin Wu, J.O. Pontt, S. Kouro, “Multilevel voltage-source-converter topologies for industrial medium-voltage drives,” IEEE Trans.Ind. Appl., vol. 54, no. 6, pp. 2930–2945, Dec. 2007

[6] Y. Patel, D. Pixler, A. Nasiri, “Analysis and design of TRAP and LCL filters for active switching converters,” IEEE industrial electronics international symposium (ISIE), pp 638-643, 2010

[7] M. Liserre, F. Blaabjerg, and S. Hansen. "Design and Control of an LCLfilter based Three-phase Active Rectifier," Conf Rec. of 36th IAS Ann. Meeting, Chicago, 2001, pp. 297-307.

[8] M. Loserre, A. Dell’Aquila, F. Blaabjerg, “Genetic algorithm-based design of the active damping for an LCL filter three phase active rectifier,” IEEE Trans. Power electronics., vol. 19, no. 1, pp. 76–86, Jan. 2004.

[9] V. Blasko and V. Kaura, “A novel control to actively damp resonance in input LC filter of a three-phase voltage source converter,” IEEE Trans.Ind. Appl., vol. 33, no. 2, pp. 542–550, Mar./Apl. 1997.

[10] Y Lang; X Zhang; D Xu;, S. Hadianamrei; H Ma, “Nonlinear

Feedforward Control of Three-phase Voltage Source Converter,” IEEE Industrial Electronics., vol. 2 pp. 1134-1137, Jul 2006

[11] R. M. Tallam, R. Naik, M. L. Gasperi, T.A. Nondahl, Hai Hui Lu, Yin Qiang, “Practical issues in the design for AC drives with reduced DC link capacitor,” Conf Rec. of 38th IAS Ann. Meeting, vol. 3, pp. 1538-1545, 2003

[12] Miao Zhixin, Fan Lingling, D. Osborn, S. Yuvarajan, “Wind farm with HVDC delivery in inertial respose and primary frequency control,” IEEE Trans. energy conversion. vol. 25, no. 4, pp. 1171–1178, Dec. 2010

3499