2010 International Conference on Power System Technology

1

A Novel Digital Automatic Voltage Regulator for Synchronous Generator
Weilin Li, Xiaobin Zhang , Huimin Li

Abstract-- In this paper, a Digital Signal Processor (DSP) based Automatic Voltage Regulator (AVR) for synchronous generator is proposed. Analysis and development of the proposed regulator is done by both computer simulations and DSP implementation. Steady state and dynamic control performance of the regulator are investigated on different working conditions (such as voltage establishment, load connection and disconnection). A hardware platform of the AVR which uses DSP TMS320F2812 as the microprocessor is set up. Several novel methods have also been adopted in the configuration of the hardware for a better dynamic performance. Experiment results show good consistency to the simulation results. Index Terms-- Automatic voltage control, voltage regulator, synchronous generator.

I. INTRODUCTION YNCHRONOUS generators operate as variable frequency machines in many applications. One of the applications is in more electric aircraft (MEA) power systems. Aircraft power systems are required to operate in a wide frequency range because the generators are driven by propulsion turbine. Automatic Voltage Regulator (AVR) is used to regulate the output voltage through controlling the excitation field. With the development of electronic technology, AVRs begin to emerge from analog to digital transformation since digital control is a cheaper and more reliable method than analog circuit. It also has advantages of more flexible, allowing easy implementation of more complex control strategies, and interfacing with other generator control and protective units [1]. In this paper, a DSP-based AVR is developed for variable frequency synchronous generator. A three-loop control structure is adopted [2]. The internal feedback loop controls the excitation current of the exciter in order to have larger gain and bandwidth. The external feedback loop provides the basic voltage regulation. An extra disturbance-rejection loop is adopted here to compensate for the changes in output voltage caused by the generator load conditions. Analysis and development of the proposed controller is
W. Li is with the Institute for Automation of Complex Power Systems, E.ON Energy Research Center, RWTH Aachen University, Aachen, 52074 Germany (e-mail: wli@eonerc.rwth-aachen.de). X. Zhang is with the Department of Electrical Engineering, Northwestern Polytechnic University, Xi’an, 710072 China (e-mail: dgl907@126.com). H. Li is with the Department of Electrical Engineering, University of South Carolina, Columbia, SC, USA (e-mail: li80@cec.sc.edu).

S

done with Saber simulator prior to the implementation with DSP. Saber simulator is developed by Synopsys company, which can use models represented by MAST modeling language to perform simulation analysis [3][4]. The interaction of the MAST language and the Saber simulator provides a powerful method of simulating systems containing a wide variety of models. Because MAST language has the features and capabilities of an Analog Hardware Description Language (AHDL), it is suitable for the diverse and complex modeling requirements of simulation in general. The mathematic model of synchronous generators is set up here. Then based on the Saber simulator, the simulation model of the synchronous generator is built, and the effectiveness of control scheme is verified. Both the steady and dynamic control performance of the voltage regulator is analyzed during different working conditions, such as voltage establishment, load connection and disconnection. Simulation results prove that the proposed solution is effective in all these conditions. A hardware platform of the AVR is also set up. DSP TMS320F2812 is used as the microprocessor. Two types of filters are adopted because of the harmonics and noise in the real circuits. One is implemented in hardware platform as antialiasing filter. The other is realized by software in order to improve the measured magnitudes. In the DSP controller, an inverse-Chebyshev filter is defined via programming. And a novel excitation circuit is also developed to improve the dynamic performance. The rest of the paper is organized as following: In section II, the overall system structure and generator model are briefly described. The control strategy is illustrated in section III. The simulation results of both steady state and dynamic state are presented and analyzed in section IV. A hardware platform is set up and the experimental results are portrayed in section V. The conclusion is made with a discussion of future research consideration in section VI. II. SYSTEM STRUCTURE AND GENERATOR MODEL In this paper, a digital AVR for a variable frequency synchronous generator is developed. The rated generator output voltage is 115V. The frequency ranges from 360 Hz to 800Hz. A typical generator in aircraft applications is a so called three-stage generator which contains two synchronous machines, a rectifier and a small permanent magnet generator (PMG), driven by the same shaft [5]. One synchronous

978-1-4244-5939-1/10/$26.00©2010 IEEE

Fig. However.2 machine is the main generator and the other smaller lower rating machine with its field winding on the stator works as the brushless exciter. For the salient pole machine. The armature winding of the exciter is on the rotor and connected to a rotating diode rectifier bridge that feeds the field winding of the main generator. There are two main parts in synchronous generators: the stator and the rotor. The PMG on the shaft serves as the power supply for excitation. the original motor equation has to be used in the quantitative analysis. Structure of three-stage generator 0 ra 0 0 ⎤ ⎡id ⎤ ⎡ pψ d ⎤ ⎡ −ωψ q ⎤ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎢ 0 ⎥ × ⎢iq ⎥ + ⎢ pψ q ⎥ + ⎢ωψ d ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ rf ⎥ ⎦ ⎣i f ⎦ ⎣ pψ f ⎦ ⎣ 0 ⎦ (3) Flux equation of the exciter: In order to set up the synchronous generator model in Saber simulator. Then the mathematical model of exciter can be described as: Voltage equation of the exciter: ⎡u d ⎤ ⎡ ra ⎢ ⎥ ⎢ ⎢ uq ⎥ = ⎢ 0 ⎢u f ⎥ ⎢ 0 ⎣ ⎦ ⎣ Fig. Winding flux is used as state variable. the speed and load of generators changes dramaticaly. CONTROL STRATEGY AND SIMULATION MODEL SETUP A. Here. which is difficult to be analyzed in Saber simulator. since there is no access to the rotor of the combined machine. Fig. 2). ⎡ ud ⎤ ⎡ r ⎢u ⎥ ⎢ 0 ⎢ q⎥ ⎢ ⎢ u0 ⎥ ⎢ 0 ⎢ ⎥=⎢ ⎢u F ⎥ ⎢ 0 ⎢u D ⎥ ⎢ 0 ⎢ ⎥ ⎢ ⎢uQ ⎦ ⎥ ⎢ ⎣ ⎣0 0 r 0 0 0 0 0 0 r 0 0 0 0 0 0 rF 0 0 0 0 0 0 rD 0 0 ⎤ ⎡ − ia ⎤ ⎢ ⎥ 0⎥ ⎥ ⎢ − ib ⎥ 0 ⎥ ⎢ − ic ⎥ ⎥⎢ ⎥+ 0 ⎥ ⎢ iF ⎥ 0 ⎥ ⎢ iD ⎥ ⎥⎢ ⎥ rQ ⎦ ⎥⎣ ⎢ iQ ⎦ ⎥ ⎡ pψ d − ωψ q ⎤ ⎢ ⎥ ⎢ pψ q + ωψ d ⎥ ⎢ pψ ⎥ 0 ⎢ ⎥ ⎢ pψ F ⎥ ⎢ ⎥ p ψ D ⎢ ⎥ ⎢ pψ Q ⎥ ⎣ ⎦ (1) Flux equation of main generator: ⎡ Ld ⎢ ⎡ψ d ⎤ ⎢ 0 ⎢ψ ⎥ ⎢ 0 ⎢ q⎥ ⎢ ⎢ψ 0 ⎥ ⎢ 3 M ⎢ ⎥ = ⎢ 2 aF ⎢ψ F ⎥ ⎢ ⎢ψ D ⎥ ⎢ 3 M ⎢ ⎥ ⎢ 2 aD ⎢ ⎣ψ Q ⎥ ⎦ ⎢ ⎢ 0 ⎣ 0 Lq 0 0 0 3 M aQ 2 0 0 L0 0 0 0 M aF 0 0 LFF M DF 0 M aD 0 0 M FD LDD 0 0 ⎤ M aQ ⎥ ⎥ ⎡ −id ⎤ 0 ⎥ ⎢ −iq ⎥ ⎥⎢ ⎥ ⎢ −i0 ⎥ 0 ⎥⎢ ⎥ i ⎥ ⎥⎢ F ⎥ 0 ⎥ ⎢ iD ⎥ ⎥⎢ i ⎥ ⎣ Q ⎥ ⎦ ⎥⎢ LQQ ⎥ ⎦ (2) There is no damping winding in exciter. 1. it is ready to set up the generator model in MAST language. the rotor will usually be fitted with damper winding. the original equation of synchronous motor is a multi-variable equation. 2. The stator is equipped with armature windings. The mathematical model of the main generator is shown as follows. output variables and the state variables of the generator should be defined firstly (required by MAST coding). 1 schematically shows the system structure of the three-stage generator. Synchronous generator simulation model III. while the rotor is equipped with DC excitation windings. For this purpose. the input variables. This converter provides the controlled excitation voltage to the exciter. Output variables are three phase voltages and currents. which can be obtained through the Park transformation. Because of the uneven air-gap flux distribution along the circumference of the armature. a feedback control is adopted here. the simulation model in SABER is set up (Fig. With the above definations. For the synchronous generator used in MEA power systems. Control Structure The basic function of the AVR is to ensure the output voltage of the generator within a desired range. which makes it hard to satisfy the . in this paper. d-q-0 equations under synchronously rotating coordinate system are used. In the modern MEA power systems. there are damper windings on daxis and q-axis of the rotor. The AVR commands a dc–dc converter connected to the rectified PMG output. Such a system characteristic requires that the main generator excitation regulation be realized through the exciter field winding on its stator. Input variables for the generator are speed and excitation current. Voltage equation of the main generator: ⎡ψ ⎢ ⎢ψ ⎢ψ ⎣ ⎡ ⎤ ⎢ L dd ⎥ ⎢ 0 ⎥ = ⎢ ⎥ ⎢3 ⎦ ⎢ M ⎣2 0 L qq fa d q f M 0 L af 0 f ⎤ ⎥ ⎡ id ⎤ ⎥ ⎢ ⎥ ⎥ × ⎢ iq ⎥ ⎥ ⎢i ⎥ ⎥ ⎣ f ⎦ ⎦ (4) Upon the above analysis on the mathematical model of the synchronous generator.

The relationship between load current and excitation current can be determined by the support of simulation. 3. The transfer function of excitation current feedback loop is shown in equation (7). Simulation model of excitation current feedback loop C. Load Current Compensation Loop Load current compensation could further improve the dynamic performance of the overall system [10]. 3. The excitation current command or reference for the inner loop is generated by the sum of the regulation action iexreg from the outer voltage regulator and the disturbance rejection iexload from the feed-forward loop. 5. which makes the AVR has better performance both in steady state and dynamic state [7]. Excitation Current Feedback Loop As the excitation current contains many harmonic components. 6.3 dynamic voltage regulation requirements with only single loop feedback structure. a Low Pass Filter is added in this loop. Steady State Results At a frequency of 400Hz. And the middle loop is a feed forward loop. s +1 i fc τ2 = Kc (6) s id +1 τ1 Fig. the linear transfer function of the synchronous generator is built with model identification method based on a series of experiments [9]. SIMULATION RESULTS A. Simulation model of excitation current feedback loop is shown in Fig. given the differential component in excitation control. The middle loop is the disturbance-rejection loop that compensates for the changes in the output voltage caused by the generator load current. 5). Fig. B. 6. The summation of the calculated compensation current iexload and regulation action current iexreg is used as the reference input of excitation current feedback control loop. which generates iexreg according to the error between output voltage and Vref. It can be . Voltage Feedback Loop A typical PID controller is adopted in the voltage feedback control loop. In the load current compensation loop. The inputs of abc-dq0 Park transformation module are three-phase currents from the generator. 4. Fig. transformation from three-phase coordinates to two-phase synchronous rotating IV. an additional high frequency pole is added to reduce the system's cut off frequency. thereby weakening the high-frequency noise introduced by differential component. The transfer function of a typical PID controller is in equation (5) [8]. Simulation model of load current compensation loop Gd (s) = Kg (τ1s +1)(τ2s +1) τ3s (5) D. the output voltage and current of the generator with resistive load are shown in Fig. In order to obtain a better performance of the PID controller. thus enables the selection of the parameters in the compensation current loop. Control structure The outer loop is the feedback loop for direct output voltage regulation of the main generator. 7. Simulation model of voltage feedback loop Fig. as can be seen from Fig. Besides. Synchronous rotation angle is identified directly from the product of motor rotation angle and pole number. Considering that there is also a differential component in PID controller. The inner feedback loop controls the excitation current of the exciter. a Low Pass Filter is designed for the excitation current feedback control. Lf K e R f (1 + s) Rf (7) Gec ( s ) = s / ωec + 1 The excitation current reference iEXref for the inner loop is generated by suming the two currents: regulation action current iexreg ( from the outer voltage regulator) and the disturbance rejection current iexload (from the load current compensate loop). Vref + − iex reg + iEX ref + + Vt − iEX iex load iload stationary coordinate system is achieved by modeling with MAST language (models on the left of Fig. Thus a three-loop based controller [6] for the automatic generator voltage regulation is adopted here.

Thanks to the role of the digital voltage regulator. 10. Fig. because of the sudden decreasing of the load current caused by load disconnection. 8. A sudden connection of load leads to a sudden increase in load current. B. This situation will weaken the main magnetic flux and increase the internal impedance of the stator. an inverse-Chebyshev filter is implemented through programming. After the change is detected. As a result. 9. The other is realized by software in order to improve the measured magnitudes. the exciting current increases rapidly so that the generator output voltage is gradually restored to the given value. . This contributes to the sudden increasing of the generator output voltage. the impact of load changes to voltage regulator system will be analyzed. 8. and finally the system reached a new steady state. Output voltage of load connection and disconnection (up) and excitation current of load connection and disconnection (down) Fig. demagnetization effects and the anti-EMF of armature current are significantly reduced. which is verified by experimental results shown in Fig. Filter Design Because of the harmonics and noise in the real circuit. CPU board of the digital AVR A.4 obviously noticed from the figure that. After digital filtering. while the RMS value of the output voltage is 112V. Fig. the digital voltage regulator forces the exciting current to decrease rapidly. the output voltage of the main generator will drop. the peak value of the output voltage of the main generator is 159V. Its main component is a DSP TMS320F2812 (microprocessor). which makes the generator terminal voltage gradually restored to the given value. Output voltage and current of main generator (up) and RMS value of output voltage (down) V. Under load disconnected situation. One is implemented in hardware platform as anti-aliasing filter. In the DSP controller. 9. And the peak value of the current is 12A. Simulation results of output voltage and excitation current under load connection and disconnection are presented in Fig. HARDWARE SET UP Configuration of the hardware platform of the AVR is introduced in this section. harmonic spectrum of the signal is eliminated. exciting current contains a large number of harmonic components. Transit State Results In the rest of this section. Note that. 7. two types of filters are adopted [5]. The CPU board of the digital AVR can be seen from Fig. due to the commutation of rotating rectifier. making a sudden increase of magnetic flux of the generator.

3 0.4 -0.2 -0.5 5 0 error(%) -5 -10 -15 -0. relative error is introduced. Because the signal period can not be divided exactly by the sampling interval.2 0. The process can be described by equation(9). Input signal spectrum (up) and output signal spectrum (down) When there is a sharp increase in the output voltage of the generator.3 -0. 0. 12 shows. 11. both Q1 and Q2 are closed. This leads to a larger excitation current decline to accelerate the process of the output voltage drop. Calculation Error Analysis In this application.1 0 deta 0.2 -0.4 -0. Fig. 57.2 -0.5 Fig. Q1 is controlled by the PWM signal. Novel excitation circuit VD2 2 1 0 0 2000 4000 Freq/Hz 6000 8000 10000 Fig. then charges the capacitor C. These results could further validate the performance of the AVR besides the simulation results. Q2 is kept in conduction state.4 0. Fig.3 0.4 -0.5 x 10 -3 Experimental results are presented and analyzed in this section. As Fig. While it costs the system about 42ms to reach a new steady state in case of a sudden load disconnection. So.2 0. both data acquisition and calculation are completed by DSP.1 0. 8000r/min. the measurement error is apparently decreased by using compensation. As can be seen. Equation (8) is used to calculate the absolute error. EXPERIMENTAL RESULTS ( 9) Where △d can be obtained from T/Ts – N.5 5 4 Mag 3 2 1 0 0 x 10 4 3 Mag 4 x 10 4 Q1 VD1 C L 2000 4000 Freq/Hz 6000 8000 10000 Q2 Fig. ΔU 2 = U '2 − U 2 = Δd u 2 ( N ) − U '2 N + Δd (8) di f dt = uC + uVD1 + uVD 2 + uRL L VI.4 0. it takes about 32ms for the system to reconfigure from a sudden load connection to steady state. 13. 10. B. especially the transit caused by load disconnection.1 0 deta 0. Output voltage of load connection (up) and disconnection (down) (rated resistive load.2 error(%) 0 -0. The excitation current flows through diode VD1 and VD2. Excitation Circuit Design In order to improve the dynamic response of AVR. Thus it is necessary to compensate the measured value to make it closer to the actual value.3 -0. Error without compensation (up) and error with compensation (down) C. 12.1 0. sampling frequency (1/Ts) and the number of sampling points (N) [11]. 11 shows. in normal working conditions. a novel excitation circuit is adopted in this paper. The output voltages of load connection and disconnection with lower rated resistive load are poytrayed in Fig. it can be calculated and compensated to the RMS value of measured data. As Fig.6A) .4 0. 14. 13 illustrates the results of the output voltages under different working conditions with rated resistive load.5 -0. The size of error has a relationship with signal frequency (1/T). -0.

However. Rosado. Han. Sep. the execution control algorithms executes upon one assumption: all the local measurements from the sensors are reliable.7A) IX. In order to obtain faster dynamic performance. 13(2). Ma. Okada Kazuho. [9] ZHANG Zhuo-ran. China. advanced ac motor drives. [7] Chang Jie. “New VF-Power System Architecture and Evaluation for Future Aircraft. Energy Conversion. His research interest are aircraft power systems.” in Proc. 1994. March.” Acta Aeronautica Et Astronautica Sinica. Analysis and development of the proposed controller is done through both computer simulations and the actural implementation. and M. respectively.S.” IEEE Trans. In this situation.S. Han. Industry Applications. “Tuning a PID controller for a digital excitation control system. F. YAN Yangguang. X. E. 42–52. Guide to Writing MAST Templates. Both the steady state and dynamic control performance of the regulator is investigated on different working conditions. Ma. June. [6] Fig. 90–95. 527-539. integrated simulation. Boroyevich. power electronics and power factor correction. Schaefer. several methods have been applied in the hardware platform. F. Xi’an. BIOGRAPHIES Weilin Li received the B. Wang Anhua. “Research on Building the Mathematical Model of Voltage Regulation System in Aero-power. X. China in 2002) and Donghua University (Shanghai. CONCLUSION AND FUTURE WORK A DSP based automatic voltage regulator for voriable frequency synchronous generator is proposed in this paper. Hughes. 2005. China. 42(2). F. 2003. 2004. Virginia Polytechnic Institute and State University. then the controller parameters and characteristics are determined from the models. McGraw-Hill. C. Guide to Writing MAST Templates. S. 2004. M.C. She is currently working toward the Ph. in 2005) respectively.” American Control Conference. Simulation and experimental results show a significant performance improvement compared with conventional regulators. 2006. 8000r/min. the controller outputs wrong control signals according to the bad sensor data. Huimin Li received the B. once the sensor is failed.S and M. He is now a Full Professor and the Director of the Institute for Airctaft Power Systems. and D. 2008. He is currently working toward the Ph. even though the real system is still in good condition. Boroyevich. “Model-based digital controller for a variable frequency synchronous generator with brushless exciter. detecting the failed sensors and rebuilding the data after sensor failure is necessary.M. In order to solve this problem. Germany. book 1. Xi’an. pp. 1998. China. Aerospace and Electronic Systems. 2005. 1. June. a catastrophic consequence on the whole system will occur.S. respectively. RWTH Aachen University. Wang. For example. 14. in 2007 and 2009. 2003. pp. better damping and stability. “Digital generator control unit for synchronous brushless gererator. USA. [10] Ghazizadeh. Northwestern Polytechnic University.S. 36th IEEE Power Electron. Wang. Xiaobin Zhang received the B. and smart grid.D degree in electrical engineering with the Institute for Automation of Complex Power Systems.” IEEE Trans. April. C. degrees in electrical engineering from Hefei Technology of University (Hefei. His research interests are protection in medium voltage DC (MVDC) power system. 93. [8] Kim Kiyong. [11] Kuroe Yasuaki. VII. Specialist Conf. VIII. Kundur. and M. CHEN Zhi-hui.” IEEE Trans.ON Energy Research Center. Rosado. S.S. and will be part of the future work. Her research interests are sensor failure detection and information rebuilding for power system. 23. “Analysis of Floating-Point Quantization Errors in Digital Control Systems: Influence of The Order of Arithmetic Steps in Controller.D degree in electrical engineering with the department of Electrical Enginnering in University of South Carolina. 1992. 41(2).” IEEE Trans. Power System Stability and Control. China. “A generator transfer function regulator for improved excitation control. The mathematic models of synchronous generators are studied first. degrees in electrical engineering from Northwestern Polytechnical University.. Power System. Xi’an.S. and smart grid. and D. “Model-based digital controller for a variable frequency synchronous generator with brushless exciter. Output voltage of load connection (up) and disconnection (down) (lower rated resistive load. degrees in electical engineering from Northwestern Polytechnical University. in 1983 and 1986. March. May. 11–14. NO.6 Ma Xiangfei. R. YANG Shan-shui. Aachen. REFERENCES [1] [2] P. VOL. May.” Master Thesis. pp. Columbia. book 2. [3] [4] [5] . SC. Anhui.

Sign up to vote on this title
UsefulNot useful