ELECTROMOTION 2009 EPE Chapter Electric Drives Joint Symposium, 1-3 July 2009, Lille, France

Dynamic Performance Analysis of a LC-Filter Grid-Connected Gas Turbine under Voltage-Sag Operation
F. Salha*, F. Colas* and X. Guillaud**
*L2EP, ENSAM CER de Lille, Lille, France **L2EP, Ecole Centrale de Lille, Villeneuve dAscq, France
Abstract The aim of this work is to analyze the dynamic behavior of a distributed generator based on gas micro turbine GMT using a grid-connected inverter with an output LC filter. In this paper, we focus, on what happen in case of disturbance on the network which induces a voltage sag on the output LC filter terminal. The proposed control strategy depends on the control of the output LC filter voltage with a resonant controller. We use a state feedback structure and pole assignment method to design the controller. Dynamic analysis was realized in case of a voltage sag operation caused by a three phase balanced default. The proposed control strategy is based on modification of the overall control system from a voltage control to a current control during a voltage sag occurrence. Real time simulations results have been discussed in the last part of this paper.

I.

INTRODUCTION

In the last decades, electricity market has seen liberalization, and a rapid growing of distributed generation sector. This is due to the trend for a large integration of renewable energy in electrical power systems and then this provides the need for a highly reliable good-quality electrical power generation. In power system with distributed energy, the low voltage distributed energy sources, such as gas micro turbine, photovoltaic system, wind turbines, storage devices, such as flywheel super capacitor, batteries, and controllable load. Form an energy system structure named microgrid. Thus, the microgrid can operate either in parallel with the utility grid or in a stand-alone mode creating an islanding system. In fact, in standalone mode operation, the loads in the microgrid can receive power from local micro sources, depending on the customers situation [1]. Typically, the micro grid must have at least one voltage generator which is responsible to regulate microgrid voltage and frequency during stand alone operation mode. Considering that the primary energy introduced to the gas micro turbine is controllable, this allows micro turbine to operate as a voltage generator, while other sources in the microgrid are operating as generator current. In this paper, the GMT has been adapted to be as

a voltage source by using three-phase voltage inverter equipped by an output LC filter. Inverter voltage control system has been established by using a resonant controller, since; it proposes a new stationary reference frame control method that introduces a cosine transfer function with a specified resonant frequency into the voltage compensator. Hence, this controller is generally made up of a proportional and a resonance term, which contains two imaginary poles that aimed to obtain an infinite gain at the resonance frequency. One of the most important problems which happens continuously in transmission and distribution systems and affects the system stability is the voltage sag. Many reasons are the origin of this problem, such as short circuit, transformer energizing, capacitor bank charging etc. During a voltage sag event, the amplitude of effective load voltage decrease [2] from 0.9 of the nominal load voltage to 0.1 in a very short time. This paper focus on the dynamic comportment, for voltage generator-based gas micro turbine in a microgrid connected to distributed network, caused by a voltage sag of 80% of the nominal value resulting from a balanced three phase short circuit. This paper is divided into four main sections. Section II describes the voltage source distributed generator based gas micro turbine with attached elements used for the connection with the distribution network. In section III, the proposed control strategy will be discussed in details. The state feedback structure is applied firstly on a single phase system and then a three phase LC filter. In section VI, analysis of system behaviour in case of voltage sag caused by a three phase balanced short circuit will be discussed. Finally, system implementation on a real time simulator will be presented and results will be demonstrated. II. VOLTAGE SOURCE DISTRIBUTED GENERATOR-BASED GAS MICRO TURBINE Generally, this model of generator consists of three main parts. Gas Micro Turbine (GMT), Permanent Magnet Synchronous Machine (PMSM), and finally power electronic converter with LC filter (see Fig. 1)

978-1-4244-5152-4/09/$26.00 2009 IEEE

Heat recuperator Heat w ater netw ork

P erm anent M agnet Synchronous M achine H igh speed Air flow

Natural gas

Turbine

Gas Compressor

Variable frequency

Combustion chamber

P M SM side converter

Grid side converter

u
Temperature Currents Currents DC voltage

D igital Power Controller

Voltages

Speed

GRiD

PW M

PW M

AC Filter 50 Hz

On the grid side, the current and voltage are controlled by action on the inverter. Considering that the power demand is always within the capability of the device and the primary generator control keeps the DC-bus voltage constant, therefore the analysis can be limited to the inverter control. In addition there is no need of representing the more complex dynamic behavior of the generation system [6]. For this purpose, we assume that an inverter interfaced with distributed generator source (GMT) is represented by a DC voltage source placed before the inverter as shown in Fig. 3.
Vbus

Inverter

LC Filter Filtre LC Grid impedance

Grid

ref

P grid_ref

Q grid_ref

Fig. 1 Block diagram of generator-based gas micro turbine [3]

The gas micro turbine considered as a source of mechanical power, is the capstone C30 [3], electrical power of 30 kW and speed more than 50 000 rpm a massive flow of air equal to 0.129 (kg/s) and a maximal torque equal to 5.72 Nm. It may be divided into three components: the compressor, combustion chamber and turbine, adding a heat recovery that takes advantage of the increased temperature of exhaust gas to heat the air entering the combustion chamber; this process aims to accelerate the process of combustion, thereby reducing the gas consumed by the turbine, A first order transfer function is used GMT modeling. References [3-5] present detailed GMT components model. The studied PMSM is a 2 poles machine with a no salient rotor. It can deliver an active power of 30 kW at a 96 000 rpm rotation speed when the voltage is 400 V. In fact, because of the highfrequency output voltage of the machine (PMSM) and the requirement to have a constant frequency (50 Hz) at the point of connection to the network, a power electronics system is used for achieving this goal. It consists of two main elements, a rectifier, which guarantees the constant frequency current to the DC bus and the inverter that allows deliver, at the point of common coupling a 50 Hz frequency voltage network. Both components use PWM technique modulation. Several loops are involved in the control of the system. On the PMSM first, the torque control is used to keep a constant DC voltage. Then the gas entering to the combustion chamber is regulating the speed rotation. Fig. 2 shows a block diagram of a GMT control system.

GMT
PMSM Machine PMSM synchrone redresseur Rectifier

Fig. 3 Model of a generator connected to the network through a voltage inverter

III. CONTROL OF A THREE-PHASE INVERTER WITH LC FILTER: STATE FEEDBACK STRUCTURE The overall control system of the inverter with an output LC filter is summarized in Fig. 4. This control strategy aims to generate a voltage, at the output of LC filter, identical to the reference voltage signal. In fact, voltage control was established by mean of a resonant algorithm.

DC Source

Fig. 4 Scheme of LC filter voltage control

The control method is based on state feedback for state variables which are LC filter, current and output voltage, and the two state variables of the controller itself. State models for both LC filter and resonant controller must firstly be formulated. A. LC Filter modeling A state model of a single phase LC filter is carried out by using the voltage and current equations depending on the state variables that are the inductance current is and capacitor voltage uc (Fig. 5).
is
Ls , Rs

im _ out _out
us

Q0 _ ref
Gas flow

control

is

Pm_ ref
Speed Boucle control Vitesse

PMSM Contrle Touque couple control

ir

us
Contrle voltage Bus continu
control DC bus

us _ ref

ic um
C

Pem _ ref

uc

ref
Fig. 2 Block diagram for GMT control system structure Fig. 5 Single-phase LC filter

u m u c = i s .R s + L s .
du is ir = C c dt

di s dt

(1) (2)

The control variable is the input filter voltage um as denoted by (3):

d dt Rs i s L s u = c 1 C 1 1 L s i s . + L s u c 0 0 .u m 0 + 1 .i r C

Firstly, we design the control for a single phase LC filter by calculating the matrix vector of gains (K) for state variables of both controller and filter Xc and Xp respectively. Then the mathematical matrix which describes the previous system denoted by state form is
d X p Ap Bp K p Bp Kc X p 0 0 + = Ac X c Bc y pref dt X c BcC p

(8)

(3)

where state and command matrix and the state-variable vector are, respectively,
Rs Ls Ap = 1 C 1 Ls , 0 1 B p = Ls 0 , X i s = u c

(4)

where p index refers to the process (LC filter) and c to the controller. As the transfer function of LC filter is of second order, that means we have two poles, and resonant controller has the same number of poles. Then, state feedback gain matrix is determined in such a way that 4 conjugated closed-loop poles have the same real part. This is realized via the pole assignment technique, as in [9], which aims at designing the feedback controller by placing the poles of the closed-loop system at desired locations as shown in Fig. 7.

A three-phase model is derived from this single phased model using coupling and decoupling matrix as explained next. B. Resonant controller modeling Many structures of resonant controller has been already used for current control as in [7]-[8]-[9], this controller has a proportional and cosine transformed resonant element. The transfer function is described by: s (5) G( s) = k + k
p i 2 s 2 + 0

Fig. 7 Pole assignment method

In order to use this control algorithm in a state feedback control, state representation of this controller is proposed: 2 (6) d y 0 p y 1
= dt u 1 + e 0 u 0

Then, the coefficients of controller can be identified from the characteristic polynomial of the closed-loop transfer function of the system by a criterion polynomial as in this equation:
P(s) = [(s a + jbi )(s a jbi )]
i =1 n

Where state and command matrix and the state variable vector are respectively: 2 0 p (7) 1 y
Ac = 1 , Bc = , 0 0 Xc = u

; {a, bi R; i, n N} (10)

with y and u, the state variables, and e, the control variable. C. State feedback control- poles placement approach When the state variables of a system are known (either through measurement or estimation), it is possible to feed their values back to the inputs with a controller in order to determine the system dynamic, this is the principle of the state feedback structure [9]. State feedback voltage control is realized for single-phase equivalent system as depicted in Fig. 6.
Yp_ref

where a determines the dynamic response and the control system stability. Pole placements were chosen in such a way that the closed dynamic has two double poles. For this purpose, the angle , mentioned in the above Fig. 7, corresponds to damping ratio of 0.7. D. Application to a thre- phase system The designed controller for the single phase system is also applied to the three phase system. This will be achieved by modifying of the system representation using coupling and decoupling matrix shifting. Fig. 8 shows three-phase the LC filter used a delta configuration for capacitors connection, denoting to state variables on it. As indicated in single phase model the current and voltage equations are written by the two next equations:

+ -

Xc e dX c = A.Xc +B .e c c dt

Kc

up

dXp

dt

= Ap.Xp +Bp.up

Xp

Yp =Cp.Xp

Yp

Kc

Fig. 6 State feedback voltage control scheme

Fig. 8 Three-phase LC filter

R d is1 = s Ls dt is 2

is1 1 2 3 + i L 1 3 s2 s

1 3 u m1 u C1 2 3 u m 2 u C 2

(11) (12)

d uc1 2 / 3C 1 / 3C is1 ir1 = dt uc 2 1 / 3C 2 / 3C is 2 ir 2

Depending on these above equations, we define:

vm1 2 / 3 1 / 3 u m1 v = 1 / 3 2 / 3 u m 2 m2 C '= 3 C vc1 2 / 3 1 / 3 u c1 = v 1 / 3 2 / 3 u c 2 c1

The current limiter, which is a PI controller, limits the inductance LC filter current during the voltage sag to a reference imposed value. In order to design the current limiter, controller parameters have to be calculated depending on the diagram in Fig. 10, where 1 + .s defines the transfer function of C (s) = K p .s controller, and 1 , the transfer function of L filter. H ( s) =
Ls .s + Rs
+ -

It comes:
d L s i s1 + R s i v v c1 dt s1 = m1 d i v v c 2 Ls i s 2 + R s s 2 m 2 dt

dv C ' c1 = i s1 i r1 dt dv c 2 C' = is 2 ir 2 dt

y ref

e (s )

C (s )

H (s )

In a similar manner as in single phase LC filter modeling, state equations for the two new phases are formulated in these two state equations:
Rs d is1 Ls = dt vc1 1 C' 1 1 0 Ls is1 . + Ls .vm1 + 1 .ir1 vc1 0 C' 0

Fig. 10 Basic structure of LC-filter current control

(13)

This leads to two decoupled set of equations for phase1 and the same for the phase 2. Then it is possible to apply the same feedback control on both decoupled systems. This control strategy has been carried out by Matlab simulinkTM. IV. BEHAVIOUR IN CASE OF VOLTAGE SAG

In a classical L or LCL grid-connected converter, the control aimed at regulating the grid current. In case of voltage sags, the current is naturally limited. In a LC grid converter, which is usually used in isolated mode operation, the output voltage is controlled. In case of voltage sag, the converter current tends to increase to support the output voltage level as in [6]. In this paper, a specific control strategy has been deduced to maintain the current in an acceptable level. This last strategy consists in modifying the overall control system, from a voltage control to current control during a voltage sag occurrence (Fig. 9).

System start-up was done without GMT generator connection to the network in order to allow the PLL system synchronization to recover the network angle. This angle is necessary for GMT generator reference voltage determining. Also, switching operation takes into account that at start-up operation and without the existing of voltage sag, GMT works as a voltage generator (vallimiter=0 and switch on position 1, Fig. 9). The next stage is during the voltage sag, where the switch disconnects the above controller and then connects the current limiter (vallimiter=1 and switch on position 2, Fig. 9) when the current exceeds a maximum current value. After the voltage sag clearing, switching operation will return to voltage and current control mode. Switching between the two controls strategies was realized by state flow and illustrated in Fig 11.

Fig. 9 Block scheme for control strategy in case of voltage sag

Fig. 11 Voltage-sag detection algorithm (Current limiter validation flowchart)

V.

REALIZATION AND DISCUSSION

Control strategy explained above was constructed on Matlab-SimulinkTM, then, it was implemented on real time simulator. Synchronization system based PLL was established in order to synchronize the two connected voltage sources which are the main grid and GMT. For normal operation mode, since two power generators were connected by impedance, active and reactive powers may be exchanged. Thus, real power loop has been realized for this purpose depending on droop control. A. Controller implementation on the real-time simulator Overall system control was implemented on a real time simulator (Hypersim) which has a hardware and software package designed to simulate power systems networks [10]. With this real time simulator, networks can be simulated off-line on workstations with one or several processors, or in real time with parallel processors. This kind of simulation allows to simulate events at a real environment speed. Hence, this allows connecting actual equipment to a network simulated by Hypersim. On the other hand, electrical equipments tests in an actual context may be complex and expensive, for example, the achievement of an actual short circuit to analyze the protection behavior should be very expensive to carry out. Fig.12 presents the main hardware elements of the Hypersim real-time simulator which is organized around a SGI four processors computer. An external PC is connected to the simulation server. It contains the Graphical User Interface, a code generator and a visualization tool by using ScopeView software (Fig. 12).

Fig. 13 Experimental test bench TABLE 1 SIMULATION SYSTEM PARAMETERS Connecting to network inductance 5 mH Connecting to network resistance 0.4 ohm LC filter resistance (Rs) 0.1 ohm LC filter inductance (Ls) 1 mH LC filter capacity (C) 20 F Operation voltage magnitude 340 v ph-g DC bus voltage 760 v DC bus capacity 5 mF a complex pole real part 2000 b complex pole imaginary part 1400 Limit Current value 28 A

Fig. 12 Real-time simulator interacting with an external electronic board

In this paper, the control strategy was implemented on real time control-command board of type dSPACE DS1103. This board is connected to another computer controlled by dSPACE package software. The interface between these two environments is necessary for testing the control in closed loop. This was realized by using IP module (Industry Pack) (Fig 12). Fig. 13 depicts the experimental test bench. It consists of two real time environment, one (Hypersim) is for the implementation of power components, as GMT, grid converter, DC link, LC filter, and the main Grid, the other DSPACE, contains the control strategy in the both faulted and normal operation modes. The studied system is that depicted in Fig. 1. System parameters are shown in Table 1. PI controller was used in GMT speed, PMSM torque, and DC bus voltage control. The simulation was carried out by GMT not connected to the distributed network; the connection was achieved at 0.5 s and the voltage sag (Fig. 14a-b) at about 18s and is cleared at 20.8 s.

The inverter control system quickly reacts, by means of state flow proposed algorithm, in order to limit the current supplied by GMT generator (Fig. 14 -c-d). Since, at the beginning of the voltage sag, the filter current exceeds the maximum value supported by the power electronic devices. Thus, current limiter was validated (Fig. 14-b). Fig. 14-d shows that the filter current was limited to the imposed value, during voltage sag. The imposed current value, during the voltage sag, was set to an amplitude of 28 A. We can notice a transitional stage on the experimental current result at moment of voltage sag occurrence (Fig. 14-d). In fact, the using of digital hardware in parallel generates a time delay which prevents the PLL to synchronize perfectly. The dynamic behavior of GMT was illustrated at its participation to support DC bus. DC bus voltage and angular speed controller were set to response times of 500 ms and 100 ms, respectively (Fig. 14-e). With this figure, we can easily find that DC link voltage has followed the reference within the set response times. The same result was shown for the GMT angular speed (Fig. 14-f). Since, there are two transitional stages on the angular speed due to the modifying of control strategy at the moment of voltage sag and at its clearance. VI. CONCLUSION The dynamic behavior of inverter interfaced GMT generator embedded on a distribution system has been studied. Power converter connection to the network was achieved by using a LC current and output voltage resonant controller, and a state feedback control strategy.

(a) Network voltage

(f) Gas micro turbine angular speed Fig. 14 Experimental results

Furthermore, control designed by the pole assignment technique was used for controller design. A solution for a voltage sag problem has proposed. For this last solution, a current limiter mechanism depending on PI controller was introduced. Different experimental results achieved by real time simulator were presented. For the future work, a strategy has to be adapted for determining the current value through voltage sag operation.
(b) LC-filter terminal voltage

REFERENCES
[1] Jaehong Kim, Junggi Lee and Kwanghee Nam Inverter-Based [2]
Local AC Bus Voltage Control Utilizing Two DOF Control ieee transactions on power electronics, vol. 23, no. 3, may 2008. H. Nasiraghdam, and A. Jalilian Balanced and Unbalanced Voltage Sag Mitigation Using DSTATCOM with Linear and Nonlinear Loads, International Journal of Electrical, Computer, and Systems Engineering (www.waset.org), Spring, 2007. P. Li, Ph. Degobert, B. Robyns, B. Franois, Modelling and Control of a Micro Turbine Generator by Using Causal Ordering Graph, Computational Engineering in Systems Applications: CESA, Beijing, China, 4-6 Octobre, 2006,CDROM. H. Nikkhajoei, M. R. Iravani A Matrix Converter Based MicroTurbine Distributed Generation System IEEE volume 20, Issue 3, July 2005 Page(s): 2182 - 2192. O. Fethi, L.-A. Dessaint, K. Al-Haddad, Modeling and simulation of the electric part of a grid connected microturbine Power Engineering Society General Meeting, IEEE 6-10 June 2004, Page(s):2212 - 2219 Vol.2. R. Caldon, F. rossetto, R. Turri Analysis of dynamic performance of dispersed generation connected through inverter to distribution networks. CIRED, 17th International Conference on Electricity Distribution Barcelona, 12-15 May 2003. N. Hatziargyriou. MICROGRIDS: Potential Allies for a More Efficient Power Supply. Key Note Speech, Medpower04, Cyprus, 2004. Marco Liserre, Remus Teodorescu, Frede Blaabjerg Stability of Photovoltaic and Wind Turbine Grid- Connected Inverters for a Large Set of Grid Impedance Values IEEE transactions on power electronics, vol. 21, no. 1, january 2006 Lpez de Heredia, H. Gaztaaga, I. Etxeberria-Otadui, S. Bacha, X. Guillaud Analysis of Multi-Resonant Current Control Structures and Tuning Methods '2006 IEEE. V. Courtecuisse, M. El Mokadem, X. Guillaud, F Salha, B. Robyns Use of real time simulation to validate primary frequency control with wind turbine IEEE PES General Meeting, Pittsburg, 20-24-july2008.

[3]

(c)

Voltage sag detection (current limiter validation)

[4] [5]

[6]

[7] [8]
(d) LC-filter current (is)

[9] [10]

(e) DC-bus voltage evolution