09-05-2012, 01:07 PM
DESIGN AND IMPLEMENTATION OF A NEURAL NETWORK CONTROLLED UPS INVERTER
[attachment=21665]
INTRODUCTION
Uointermptible power supplies (UPS) should provide a
sinusoidal voltage IO its customers with constant magnitude.
However, voltage distortion arises when nonlinear loads
mvolve, e.g., diode rectifiers, which draw non-sinusoidal
current when Fed with a sinusoidal voltage. As growing
U~IIIWUOII 01 nonlmear power electronics loads, it is
bc;oniing inox dillicult and siguficant to design a suitable
cotitroller for UPS inverter, the core of UPS, to maintain a
siiiusoidal voltage espccially for thc cases of nonlinear
loads.
Some 1nultiplefeedback-loop control strategies for the
UPS inverter were proposed in [l][2]. Such schemes sense
the current in the capacitor (or inductor) of the LC filter to
fonn an inner current Fccdback loop incorpolaung with an
outer voltage feedback loop. Owing to the introduction of
tl:? inner currcnt feedback loop, the ourput impedance of the
tnvcncr is decreased and the dynaniic stiffness enhanced.
The controlled inveller can produce a satisfactoly sinusoidal
output voltage within a ceaain load range.
DIAGRAMO F ups INVERTER AND LINEAR
MODEL
Fig. 1 shows the circuit diagram of a single-phase halfbridge
voltage sonrce UPS inverter, followed by a LC filter.
PWM generator and gate drive circuits are also illustrated.
R, represents the resistance of the filter inductor. The
effective series resistance @SR) of the filter capacitor is
ignored since it only has a small effect within the frequency
range concerned.
Because the switching frequenq & (here is 20KHz) is
several orders bigher than the fnudamental frequency of the
AC output, the dynamics of the PWM inverter can be
ignored. Thus, the UPS inverter can be modeled as a simple
propoaional gain block. Fig.2 shows a linear model of the
inverter system BWM inverter plus the output filter and the
load), in which the proportional gain of the inverter, K, is
equal to Vdc/Vc (vdc is the voltage of the DC power source
and V, is the peak voltage of the hiangnlar carrier wave).
CONCLUSIONS
A neural network control scheme for UPS inverter has
been presented in this paper. First, the methods for obtaining
the example pattern are introduced. Two simulation models
are built to obtain example patterns for linear and nonlinear
loads respectively. One is a multiple-feedback-loop
controller for linear loads; and the other is an idealized loadcurrent
feedback controller specially designed for nonlinear
loads. Then a neural network is selected to train nsing
example patterns so as to formulate the control law. The
proposed neural network controller is implemented using a
simple analog circuit. Experimental results confirm that the
circuit can function the proposed ANN controller. By
comparisons with a traditional PI controller, it is shown that
the proposed neural network controlled UPS inverter has
good steady-state and transient responses, and can decrease
the THD under nonlinear loading conditions.
[attachment=21665]
INTRODUCTION
Uointermptible power supplies (UPS) should provide a
sinusoidal voltage IO its customers with constant magnitude.
However, voltage distortion arises when nonlinear loads
mvolve, e.g., diode rectifiers, which draw non-sinusoidal
current when Fed with a sinusoidal voltage. As growing
U~IIIWUOII 01 nonlmear power electronics loads, it is
bc;oniing inox dillicult and siguficant to design a suitable
cotitroller for UPS inverter, the core of UPS, to maintain a
siiiusoidal voltage espccially for thc cases of nonlinear
loads.
Some 1nultiplefeedback-loop control strategies for the
UPS inverter were proposed in [l][2]. Such schemes sense
the current in the capacitor (or inductor) of the LC filter to
fonn an inner current Fccdback loop incorpolaung with an
outer voltage feedback loop. Owing to the introduction of
tl:? inner currcnt feedback loop, the ourput impedance of the
tnvcncr is decreased and the dynaniic stiffness enhanced.
The controlled inveller can produce a satisfactoly sinusoidal
output voltage within a ceaain load range.
DIAGRAMO F ups INVERTER AND LINEAR
MODEL
Fig. 1 shows the circuit diagram of a single-phase halfbridge
voltage sonrce UPS inverter, followed by a LC filter.
PWM generator and gate drive circuits are also illustrated.
R, represents the resistance of the filter inductor. The
effective series resistance @SR) of the filter capacitor is
ignored since it only has a small effect within the frequency
range concerned.
Because the switching frequenq & (here is 20KHz) is
several orders bigher than the fnudamental frequency of the
AC output, the dynamics of the PWM inverter can be
ignored. Thus, the UPS inverter can be modeled as a simple
propoaional gain block. Fig.2 shows a linear model of the
inverter system BWM inverter plus the output filter and the
load), in which the proportional gain of the inverter, K, is
equal to Vdc/Vc (vdc is the voltage of the DC power source
and V, is the peak voltage of the hiangnlar carrier wave).
CONCLUSIONS
A neural network control scheme for UPS inverter has
been presented in this paper. First, the methods for obtaining
the example pattern are introduced. Two simulation models
are built to obtain example patterns for linear and nonlinear
loads respectively. One is a multiple-feedback-loop
controller for linear loads; and the other is an idealized loadcurrent
feedback controller specially designed for nonlinear
loads. Then a neural network is selected to train nsing
example patterns so as to formulate the control law. The
proposed neural network controller is implemented using a
simple analog circuit. Experimental results confirm that the
circuit can function the proposed ANN controller. By
comparisons with a traditional PI controller, it is shown that
the proposed neural network controlled UPS inverter has
good steady-state and transient responses, and can decrease
the THD under nonlinear loading conditions.