2011 International Conference on Mechatronic Science, Electric Engineering and Computer August 19-22, 2011, Jilin, China

Development of a Nonlinear Control Strategy for the DC Power Transferring System

Gang YANG
Department of Power & Energy Systems Ecole Suprieur dElectricit (SUPELEC) Gif-sur-Yvette, 91190, France yjsyanggang@gmail.com

Abdelkrim BENCHAIB, Guillaume DE-PREVILLE

Power Electronics Activities Alstom Grid Massy Cedex, 91300, France

AbstractThis paper presents a detailed control strategy to a PWM bidirectional DC-DC transferring system designed for high voltage, high power applications. The objective is to realize precise bidirectional power flow control between two DC grids, where the half bridge topology is adopted for the connection between them. By using the averaging technique, the systems nonlinear average model is derived, based on which the nonlinear control algorithm is proposed, to realize the current control. The proposed circuit and control strategies are simulated using SimPowerSystems and a comprehensive synthesis of the control algorithms is conducted, from the viewpoint of rapidity, robustness and system cost. It is verified that the proposed nonlinear control algorithm has a fast response, competent disturbance rejection ability and low cost. Although the structure studied here is the half bridge bidirectional converter, however, the controller design methods and the relative conclusions can be promoted to any type of structure in power transferring systems. Keywords-half bridge converter; power flow control; nonlinear control; Lyapunov function

I.

INTRODUCTION

Power flow control between two DC grids is an area that has seldom been touched before. The objective of constructing such a system is to realize the bidirectional power flow control between two DC grids. In order to connect the two DC grids with different voltage levels, a DC-DC converter is indispensable for utilization. The bidirectional half bridge converter is proved to be an effective solution for high voltage-high power conditions, as fewer semiconductor devices are utilized. Moreover, for the purpose of eliminating the high frequency current fluctuations caused by the switchs commutation, two current filters are added to ensure that the currents debited by the two DC grids are little-rippled (with regard figure 1). However, the introduction of two current filters also increases the order of the system, thus increases the control complexity. There exist many strategies for stabilizing and regulating DC/DC converters output. The objective of controlling such proposed systems is to provide better performances, ensure closed loop stability and guarantee the adequate behavior. In terms of converter controller design, the use of averaging techniques leads to the establishment of a nonlinear mathematical model (Krein et al., 1990). This model is often followed by linearization and small-signal analysis which

allows the derivation of a linear time-invariant dynamic model for a specified converter topology, thus this approach enables the designer to use a linear controller to realize the regulation objectives (Samosir and Yatim, 2010; Tan et al., 2006). As to the linear regulator design of power electronics converters, the author (Raviraj and Sen, 1997) compares the performance of three different controllers, which are classical PI controller, the fuzzy logic controller and the sliding mode controller. However, the design of the linear regulator is obtained by the linearization analysis and thus influenced by the systems steady state, which is a main drawback of the existed linear control algorithm. Moreover, in our proposed system, the added input and output current filter makes it a 5-order system which signifies that a simple linear PI or PID controller is very cumbersome for regulation (Ghadimi et al., 2007). Under this circumstance, other researchers try to explore the non-linear characteristics of the dc/dc converters and proposed several relative control algorithms (Femia et al., 2003; Chan, 2007; Hu, 2011), but the proposed methods are too difficult to be comprehended and realized in industrial utilization. The control algorithm proposed in this paper is totally based on the systems nonlinear characteristics and is comparably easy to realize. The synthesis of the power transferring system and the associated control strategy will be discussed in detail in the following parts. II. SYSTEM MODELLING

Supposing that the power transferring system studied here realizes the bidirectional power flow between two DC grids, 150kV for Port A and 30kV for Port B, separately. The system consists of a half bridge bidirectional converter and two current filters, as shown in the figure 1.

Figure 1. The studied power transfering system

The switch K1, the diode D1 and the inductance L form a standard buck converter. Adding an anti-parallel diode D2 and

978-1-61284-722-1/11/$26.00 2011 IEEE

2627

a controllable switch K2 makes the circuit possible to operate in two directions. To realize the bidirectional power flow, the switch K1 and K2 are controlled by PWM signal in the complementary way. In order to eliminate the current fluctuations caused by the circuit switching, a current filter composed by L1 and C1 is placed at the side A and another current filter composed by L2 and C2 is placed at the side B. The resistance rL1, rL and rL2 are the series resistance of the inductance L1, L and L2, respectively. We define that when the power transferring direction is from A to B, the circuit operates in down mode; to the contrary, if the current flows from the B to A, the circuit operates in up mode. To obtain a higher control quality, the considered modeling technique must describe as precisely as possible the systems nonlinearities and take into account all of its elements. Thus before system modeling, these following hypotheses are important: (1)Neglecting the switchs and diodes forward voltage. (2)Considering the switchs and diodes on resistance, neglecting its parasite capacity. (3) Considering the inductance series resistance rL1, rL and rL2, respectively, for the three inductors. Supposing that the circuit operates in the continuous conduction mode (CCM) and the power transferring direction is from the port A to the port B (down mode). In this mode, there are apparently two operational configurations, which depend on the position of K1. When K1 is switched on, it is defined as the configuration A; when K1 is switched off, it is defined as the configuration B. A. Down mode, Configuration A Between the phase [0, dT], K1 is switched on, the diode D1 is reverse-biased, the systems equivalent circuit is shown in the following figure.

diL1 ve = iL1rL1 + L1 dt + vc1 i = C dvC1 + i 1 L L1 dt diL + vc 2 vc1 = ( rK 1 + rL )iL + L dt dvC 2 iL = C 2 dt + iL 2 v = v + r i + L diL 2 s L2 L2 2 C2 dt

(1)

Therefore, the state space model at the interval [0 dT] can be described by the following equation (the objective is to regulate the output current iL2 thus it is considered as the output in the state space model):

x = A1 x + B1ve + C1vs y = iL 2 = E1 x

(2)

B. Down mode, Configuration B Then at the phase [dT, T], K1 is switched off, D1 is forwardbiased, thus the system equivalent circuit can be shown in the Figure 3.

Figure 3. The equivelent circuit of the configuration B in down mode

Similarly, the following equations can be obtained.

Figure 2. The equivelent circuit of the configuration A in down mode

Selecting the inductors currents and capacitors voltages as the state variables, x=[x1 x2 x3 x4 x5]T =[iL1 vC1 iL vC2 iL2]T, the following equation can be obtained.

diL1 ve = iL1rL1 + L1 dt + vc1 i = C dvC1 1 L1 dt diL + vc 2 0 = (rD1 + rL )iL + L dt dvC 2 iL = C2 dt + iL 2 v = v + r i + L diL 2 s L2 L2 2 C2 dt

(3)

Therefore, the state space model in the interval [dT, T] can be described by the following equation:

x = A2 x + B2ve + C2vs y = iL 2 = E 2 x

(4)

The conventional way for modeling power converters is to take an average value of the system state over a switching cycle (Guesmi et al., 2008). In order to obtain this model, we consider that the converter dwells between its two configurations, thus the systems state space average model

2628

can be expressed by grouping the two state space equations using the duty cycle d, shown as in the equation (5):
d x = A x + B ve + C vs = dt [ A1d + A2 (1 d )] x + [B1d + B2 (1 d )] ve + [C1d + C 2 (1 d )] vs y = iL 2 = E x = [E1d + E 2 (1 d )] x

0 X2 = 1 L2

1 x3 C2 x4 01 vs + C2 + r L 2 x5 L 2 0 L2
x . = 4 x5

(7)

(5)
rL1 L1 1 Where C1 A = 0 0 0 1 L1 0 d L 0 0 0 m 1 C2 0 d C1 0 0 0 1 L2 1 L 0 , with 0 0 1 C2 rL 2 L2

x1 Where the X = x , X 1 2 2 x3

In the equation (6) and (7), we use the simplified form x to represent the state space average vector <x>, for the purpose of briefly illustrating the system and the control strategy. But the x used in this part also signifies the averaged vector obtained in a switching cycle. The nonlinear regulator is designed step by step, which means that we begin from the sub-equation which contains the reference signal and try to find out the variation of the other variable. The controller is obtained directly from the state space nonlinear average model, which contains some nonlinear parameters. It has been presented a nonlinear adaptive controller for the regulation and synchronization of interleaved buck converters and has obtained good transient response and closed loop stability (Fadil and Giri, 2009). But the described controller can not fully adapt to our proposed system. A new nonlinear control algorithm will be explored specially for the power transferring systems descried in this paper. 1) Take firstly the last sub-equation of the equation (7) into consideration, as shown in the equation (8). 1 1 r (8) vs x5 = x4 L 2 x5 L2 L2 L2 The objective of systems regulation in this part is to find a control algorithm that permits to get the variation of x4 for regulating x5 to x5ref (iL2ref). With the help of Lyapunov function, which is shown in the equation (9), it is possible to design such a regulator.

1 r +r r r m = D1 L K 1 D1 d ; B = L L L1
T

0 0 0 0 ;

T 1 C = 0 0 0 0 ; E = [0 0 0 0 1] . L2

x denotes the system states that have been averaged

over one switching cycle. Noting that in the up mode, the power transfers from Port B (30kV) to Port A (150kV), however, the change of power transferring direction does not affect the two operational configurations of the circuit. The equivalent circuits established in the up mode are identical to those in the down mode (if there are no significant differences between the on resistance of switch rK and that of the diode rD). The modeling result for the up mode remains the same as the equation (5) thus there is no need to remodel the system once again for the circuit operating in up mode. III. CONTROL ALGORITHM

Vx 5 =

1 ( x5 x5 ref ) 2 2

(9)

By incorporating the equation (8) to the equation (9), the following equation is obtained:
Vx 5 = ( x5 x5ref )(x5 x5 ref ) = E x 5 ( r 1 1 x4 L 2 x5 vs x5 ref ) (10) L2 L2 L2

A nonlinear control algorithm that fully adapted to the systems nonlinearities characteristics is proposed to regulate the systems output, for the purpose of providing more alternatives and comparing the performances of the two control strategies. For the state space average equation (5), the model is reassembled to separate the linear and nonlinear part .The equation can be written in the following form:
rL1 L1 1 X1 = C 1 0 1 L1 0 0 0 1 0 L x1 0 (6) x + x3 d + 01 v + 0 0 2 e C1 x4 0 x3 x R R L 1 2 2 x3 L L L

Where the E x 5 is the regulating error. Achieving the tracking objective means forcing the error to vanish. To ensure the stability, it has to maintain that the:

Vx 5 0
Supposing that :

(11)

1 r 1 x4 L 2 x5 vs x5 ref = x 5 E x 5 x 5 sign( E x 5 ) (12) L2 L2 L2 With parameters x 5 > 0 , x 5 > 0 . It can be deduced that:

V x 5 = E x 5 [ x 5 E x 5 x 5 sign ( E x 5 ) ] = x 5 E x 5 x 5 sign ( E x 5 ) E x 5 0
2

(13)

2629

Thus, we have the control algorithm developed for the model (8), which is shown in the equation (14).
r 1 x4 = x4 ref = L2 L 2 x5 + vs + x5 ref x 5 E x 5 x 5 sign( E x 5 ) (14) L2 L2

Vx 3 =

1 ( x3 x3ref ) 2 2

(23)

The time-derivative of Vx3 is:

Thus, theoretically, the equation (14) is developed for regulating the sub-system (8) to get the x4ref for a given desired x5ref. The sufficient condition for assuring the stability is to assure x 5 > 0 and x 5 > 0 . In fact, regulating the coefficient

x5

permits to ameliorate the response time and regulating the

(24) R1 1 1 x3 + ( x 2 R2 x3 )d x 4 x3ref ) L L L Where the E x 3 is the regulating error. To ensure the stability, it has to maintain that the: = E x3 (

V x3 = ( x3 x3ref )(x3 x3ref )

coefficient x 5 permits to assure the robustness of the closed loop system. 2) Take into consideration the other subsystem in the model (7).

Vx 3 0
Supposing similarly that:

(25)

x4 =

1 1 x5 + x3 C2 C2

(15)

With parameters x 3 that:

1 1 R1 x3 + (x2 R2 x3 )d x4 x3ref = x 3 E x 3 x 3 sign ( E x 3 ) (26) L L L

> 0 , x 3 > 0 . It can be deduced

In the same way, considering the following Lyapunov function:

Vx 3 = E x 3 [ x 3 E x 3 x 3 sign( E x 3 )] = x 3 E x 3 x 3 sign( E x 3 ) E x 3 0
2

(27)

1 Vx 4 = ( x4 x4 ref ) 2 2
The time-derivative of Vx4 is:

(16)

Thus, it is obtained the control algorithm developed for the model (22), which is shown in the equation (28).
1 1 R d ref = (x 2 R2 x3 ) 1 x3 + x4 + x3ref x 3 E x 3 x 3 sign ( E x 3 ) (28) L L L
1

1 1 Vx 4 = ( x4 x4 ref )(x4 x4 ref ) = E x 4 ( x5 + x3 x4 ref ) (17) C2 C2

Where the

E x 4 is the regulating error. To ensure the stability,

Vx 4 0
(18)

it has to maintain that the: Supposing similarly that:

The x4 is acquired from the model equation (15) and the x2 can be measured directly from the system, the obtained dref is used for the system regulation. In total, the equations (8), (15), (22) are the system models waiting for control, the equations (14), (21), (28) constitute the algorithms developed for the nonlinear control. Summarizing the above six equations, the schema of the controller can be presented in the following figure. It has to be reminded that a saturation part is indispensable to be added after the derivative part in the real simulation and application occasions to avoid the accidents caused by fast reference signal change.

1 1 x5 + x3 x4 ref = x 4 E x 4 x 4 sign( E x 4 ) (19) C2 C2

With parameters x 4 that:

> 0 , x 4 > 0 . It can be deduced

Vx 4 = E x 4 [ x 4 E x 4 x 4 sign( E x 4 )] = x 4 E x 4 x 4 sign( E x 4 ) E x 4 0
2

(20)

Thus, we have the control algorithm developed for the model (15), which is shown in the equation (21).

1 x3 = x3ref = C 2 x5 + x4 ref x 4 E x 4 x 4 sign( E x 4 ) (21) C2

Here, x5 is obtained directly from the model in equation (8). 3) Thirdly, the generated x3ref can be applied in the same way to get the the dref ,which can be used directly for the regulation process. Taking into account the subsystem in the model (6).

x3 =

R1 1 1 x3 + (x2 R2 x3 )d x4 L L L

(22)
Figure 4. The nonlinear controller design schema for the system regulation.

In the same way, consider the following Lyapunov function:

2630

IV.

SIMULATIONS AND RESULTS

This part concerns mainly the simulation of the proposed control algorithm. The electrical parameters used for simulation are shown as in the Table 1.
TABLE I. THE ELECTRICAL PARAMETERS ADOPTED FOR SIMULATION Buck converter rK1=rK2=0.01Ohm rD1=rD2=0.01Ohm rKL=1Ohm L= 0.7H Filters rL1=rL2=50Ohm L1= L2=0.07H C1= C2=0.33mF sources ve=150kV ve=10kV vs=30kV vs=1kV

(a)

(b)

Figure 6. The perturbations of input and output voltages of the DC power system

The modelling system is shown in the figure 5, which can be divided into 3 parts: the main system, the measurement and the control part. The main system is a classical half bridge bidirectional converter, which is realized by adding a diode with the IGBT or adding an IGBT with the diode. The two IGBTs are commanded in the complementary way thus a logical operator NOT is used for the switching of the second IGBT K2. This configuration guarantees the bidirectional current flow of the system between the DC grid 150kV and the DC grid 30kV. The simulation front panel is shown in the following figure:

Otherwise, the system should also be able to operate in the other power transferring direction, thus at t=0.8s, a command of iref =-50A is added. In this condition, the power flow direction should be: from the DC grid 30kV to the DC grid 150kV. The simulation results and the zooms at several instants for the current iL2 are shown below.

(a)

(b)

(c)

(d)

Figure 5. The simulation front panel using the proposed nonlinear algorithm

The simulation configurations are prescribed as follows: At the time t=0.1s, a command of iref=50A is imposed. To test the robustness of the closed loop system, the DC grid voltages are supposed to be with several permanent perturbations. At t=0.4s, a perturbation of ve (150kV160kV) is added; and at t=0.6s, a perturbation of vs (30kV31kV) is added. The variation of the input voltage and the output voltage is shown in the following figure.

(e)

(f)

Figure 7. (a)-(e): Current responses of iL2 under nonlinear control, and the zoom at t=0.1s, 0.4s, 0.6s and 0.8s. (f) The relative variation of duty cycle

The current response iL2 is a very important indicator of the power received or debited by the Port B (DC grid 30kV), at down mode or up mode, separately. As shown in the figure 7(b), the current response to the reference signal is very fast (response time of 0.4s), with an overshot of 34%. In the figure 7(c), the perturbation ve causes nearly negligible influences to the systems output, which confirms the robustness of the proposed nonlinear controller. As shown in 7(d), it is obvious

2631

that the perturbation of vs causes a current response but it is very fast and there is no steady-state error. Finally, the bidirectional power flow is realized at the time of t=0.8s, as shown in the figure 7(e). The figure 7(f) verifies that the variation of the duty cycle under nonlinear control strategy is limited to [0, 1]. The variation of the duty cycle at different stages shows clearly the power flow conditions. At the time of t=0.1s, when the reference signal of 50A is added, the duty cycle is regulated to be larger than 0.2, at approximately d=0.218. At t=0.4s, the input voltage ve is increased and thus the duty cycle is regulated a little smaller, approximately d=0.205, to maintain a current of 50A. To the contrary, at t=0.6s, the output voltage vout is increased; the duty cycle is increased to d=0.211. At t=0.8s, a reference signal of -50A is added and the duty cycle is reduced to a value below 0.2, approximately d=0.177. CONCLUSIONS This paper presents a nonlinear control algorithm for regulating the power flow between two DC grids. The proposed nonlinear control algorithm is a new approach for controlling the DC-DC bidirectional power transferring system, based on the exploration of systems nonlinearities and Lyapunov function. The system is modeled in the environment Simulink by using SimPowerSystems and the simulation results obtained for the proposed control algorithms are illustrated and explained. The proposed control shows a rapidity of about 0.1s to enter to the steady state. As to the aspect of system robustness, the nonlinear controllers performance is also competent, especially to the perturbation of the input voltage where nearly no influence at the transferring current is provoked. What is more, the traditional control algorithm is based on the systems small signal average model which is a linearization result of the original nonlinear model. During the linearization process, there are inevitably several losses of the systems nonlinearity characteristics and the obtained small signal model is not sufficient to present all the properties of the system. The proposed control nonlinear control method has successfully avoided these disadvantages. Moreover, it is simple to be

realized by DSP programming and only the signal measurement of VC1 is sufficient for regulation which reduced significantly the cost. In conclusion, the proposed nonlinear control exploits the nonlinearity and time-varying properties of the system, which make it as a superior control strategy for DC power transferring system. The above explained control algorithm can not only be applied to half bridge topology based DC systems, but also be possible to be promoted to all the other topologies.

REFERENCES
[1] Krein, P. T., Bentsman J., and Lesieutre, B. L. On the use of averaging for the analysis of power electronic systems, IEEE Transactions on Power Electronics, 1990, vol. 5, pp. 182190. [2] Samosir, A. S., and Yatim, A. H., Dynamic evolution control for synchronous buck DC-DC converter: theory, model and simulation, Simulation Modeling Practice and Theory, 2010, vol.18, pp. 663676 [3] Tan, S.C., Lai, Y.M., Tse, C.K., Cheung, M.K., and Maxwell, J.K., Adaptive feed-forward and feedback control schemes for sliding mode controlled power converters, IEEE Transactions on Power Electronics, 2006, vol.21, pp. 182192. [4] Raviraj, V.S. and Sen, P.C., Comparative study of proportional integral, sliding mode, and fuzzy logic controllers for power converters, IEEE Transactions on Industrial Electronics. 1997, vol.33, pp. 518524. [5] Ghadimi, A. A., Rastegar, H, Keyhani, A., Development of average model for control of a full bridge PWM DC-DC converter, Journal of Iranian Association of Electrical and Electronics Engineers, 2007, vol. 4, pp. 5259. [6] Femia, N., Vitelli, M., Time-domain analysis of switching converters based on a discrete-time transition model of the spectral coefficients of state variables, IEEE Transactions on Circuits and Systems,-I: Fundamental Theory and Applications, 2003, vol. 50, pp. 1447-1460. [7] Chan, C., A nonlinear control for DC-DC power converters, IEEE Transactions on Power Electronics, 2007, vol. 22, pp. 216-222. [8] Hu, T., A nonlinear-system approach to analysis and design of powerelectronic converters with saturation and bilinear terms, IEEE Transactions on Power Electronics, 2011, vol. 26, pp. 399-410 [9] Guesmi, K., Essounbouli, N., Hamzaoui, A., Systematic design approach of fuzzy PID stabilizer for DCDC converters, Energy Conversion and Management, 2008, vol. 49, pp. 2880-2889. [10] Fadil, H. E. and Giri, F., Robust nonlinear adaptive control of multiphase synchronous buck power converters, Control Engineering Practice, 2009, vol. 17, pp. 1245-1254.

2632