Frequency / Duty Cycle Current-Mode Fuzzy Control for LCC Resonant Converter Manli Hu, Joachim Böcker, Norbert Fröhleke DEPARTMENT OF POWER ELECTRONICS AND ELECTRICAL DRIVES (LEA), PADERBORN UNIVERSITY Warburger Str. 100, D-33098 Paderborn, Germany Tel: +49/(0)5251-60-3145 Fax: +49/(0)5251-60-3443 E-Mail: {hu, boecker, froehleke}@lea.upb.de URL: http://wwwlea.uni-paderborn.de Acknowledgements The authors would like to thank European Commission for funding this project. Keywords «Fuzzy control», «Adaptive control», «Resonant converter» Abstract A novel current-mode nonlinear fuzzy controller with adaptive gain is developed for the LCC resonant converter applied in very low frequency high voltage generators using switching frequency and duty cycle as control variables. The designed fuzzy controller is verified through simulations. Comparison between the fuzzy controller and a standard linear controller indicates the correctness and effectiveness of the novel approach. I. Introduction Mobile very low frequency (VLF) high-voltage (HV) generators are required for examination of high voltage cables during commissioning or services. Such a test generator has to provide the full voltage of some tens to hundreds of kV at a relatively small power of few to tens of kW. Compared with 50 Hz-voltage test systems, the VLF-generators have considerable advantages in terms of volume, weight and energy consumption [1]. The well-known LCC resonant converter is an attractive solution for such VLF HV generators due to its large voltage conversion ratio, natural soft-switching feature, preferable wide load and voltage operating region and ability to make use of parasitics such as leakage inductance, winding capacitance of HV-transformer and capacitors for controlling the static and dynamic voltage stress across HV-rectifiers [2] [3] [4]. In high-voltage DC applications, the Cockcroft-Walton (CW) multiplier is usually adopted, which shows lower losses and size as an equivalent transformer if the CW stage number is limited to 3-5 [5]. As an example, Fig. 1 shows a circuit diagram of the LCC resonant converter: the zero-voltage switched LCC circuit is connected Fig. 1 Circuit diagram of high-voltage generator with LCC resonant converter [3] 1 with a symmetrical three-stage CW multiplier via a step-up transformer. Based on the derived large-signal and small-signal models of the LCC resonant converter, a conventional linear PI controller was developed and implemented in prototypes due to its simple structure and ease of design [3] [6]. Since the LCC resonant converter is an evident nonlinear system, the conventional linear PI controller has limitations on its control performance. As an alternative, a novel current-mode fuzzy proportional-derivative (PD) controller with adaptive gain is developed for the LCC resonant converter in this paper. As another novelty, due to the wide range of output voltage and load, and the allowed operation margins of switching frequency fs and duty cycle d, both of them are employed as the control variables and integrated in the fuzzy control strategy. This paper is organized as follows: Section II briefly introduces the characteristics of the LCC resonant converter and the limitations using standard linear PI control. Section III represents a currentmode fuzzy PD control design, including the introduction of rule base, membership functions and control block diagram. Section IV shows simulation results of the fuzzy control in comparison with a conventional linear control. Section V summarizes main results of the paper and gives an outlook. II. Limitation of Linear PI Control The steady-state voltage gain of the LCC resonant converter (Fig. 1) is shown in Fig. 2. The voltage gain M depends both on switching frequency fs and duty cycle d. However, due to specified output voltage range, load range and other operating limits, the converter is operated only within the shaded operating region that is spanned by boundary operating points OP1-OP8. The voltage conversion ratio M, normalized switching frequency fsn, characteristic impedance Z and the quality factor Q of the resonant tank are also denoted in Fig. 2 (the capital letter indicates the steady-state). Among which, n is transformer turns ratio, k is stage number of CW-generator, f0 is the natural resonant frequency, and R0 is the equivalent resistor in each steady-state operating point (refer to [3] for details). The large-signal and small-signal models of the LCC resonant converter were derived in previous papers [3] [6], which are used for the design of a conventional linear PI controller. However, since the VLF HV generator is naturally a nonlinear system, the linear controller has inevitable drawbacks such as lower dynamics and variable control bandwidth for different operation points. Fig. 3 shows the loop gains of the boundary operating points with linear PI controller: the control bandwidth with frequency control for different Q is nearly the same: fn=0.01 (fn is the normalized modulation frequency regarding f0); while the control bandwidth with duty cycle control are different for various Q, the Fig. 2 Voltage gain M LCC resonant converter left: M vs. normalized switching frequency f sn = f s / f 0 right: M vs. duty cycle d: M= Uo 2 ⋅ n ⋅ k ⋅ Uin , f0 = 1 Cs ⋅ C p Ls R0 Z= Cg = 2π Ls Cg Cg Q = Z Cs + C p , , , 2 Fig. 3 Loop gains of boundary operating points with linear PI controller (left) OP1-OP3 with frequency and (right) OP5-OP7 with duty cycle control maximal crossover frequency is nearly same as fn=0.7, while the minimal is fn=0.02. It can be concluded that with such small-signal characteristics, a conventional linear controller cannot get satisfying performance for a wide operating range. Some nonlinear control scheme is expected to yield enhancement. Fuzzy logic control is successfully implemented over the past two decades in many industrial control applications and has been proven to be a successful control approach to many complex nonlinear systems. Considering the simple structure and easy design of linear control, it is reasonable to combine fuzzy logic control and conventional linear control together to design a nonlinear controller with simple structure and high efficiency. It is well known that an inner current-mode control is beneficial to cancel the gain variation of the outer voltage loop [4]. As a result, a current-mode fuzzy proportional-derivative (PD) control using switching frequency fs and duty cycle d is developed in this paper. This control approach is further supplemented by an adaptive scaling gain with respect to the reference resonant effective current in order to get improved control performance for various operating points. III. Current-Mode Fuzzy PD Control The basic structure of a fuzzy control system applied in LCC resonant converter is shown in Fig. 4. The fuzzy controller consists of four conceptual components: knowledge base, inference mechanism, fuzzification interface and defuzzification interface. The knowledge base contains all the controller knowledge and it comprises a fuzzy control rule base and a data base. The data base is the declarative part of the knowledge base which describes definition of objects and definition of membership Fig. 4 Basic structure of fuzzy control system applied in LCC resonant converter 3 functions. The fuzzy control rule base is the procedural part of the knowledge base which contains information on how these objects can be used to infer new control actions. The inference mechanism is a reasoning mechanism which performs inference procedure upon the fuzzy control rules and gives conditions to derive reasonable control actions, which is the central part of a fuzzy control system. The fuzzification interface (or fuzzifier) defines a mapping from a real-valued space to a fuzzy space, and the defuzzification interface (or defuzzifier) defines a mapping from a fuzzy space over an output universe of discourse to a real-valued space [7]. As a model-free control, the knowledge base of fuzzy control is mostly dependent on a good understanding of the behaviour of the LCC resonant converter. From [3] and [4], the low frequency gain of control-to-output transfer functions of LCC resonant converter: Gf0 (regarding fsn) and Gd0 (regarding d) are proportional to the respective slope of the DC conversion ratio curve at the given operating point. From Fig. 2, it is clear that in preferable operating region, Gf0 is negative and Gd0 is positive. Also, due to the similar change pattern of the gain of control-to-resonant-current transfer function and that of control-to-output-voltage transfer function, the following formulas can be obtained: Gf 0 ∝ ∂i ∂M ∝ L <0 ∂f sn ∂f sn (1) Gd 0 ∝ ∂M ∂iL ∝ >0 ∂d ∂d (2) Based on above principle and practical experiences, the corresponding rule base for current-mode fuzzy control is represented in Table I. The error signal e(t) is the difference between the reference resonant effective current iL_eff* and the real resonant effective current iL_eff; its derivative is de(t)/dt, denotes the change-in-error. Both of them are selected as the fuzzy controller inputs. The ranges of e(t) and de(t)/dt are divided into nine subsections and indicated as “0” to “8”, respectively. Among which, “0” corresponds to negative maximal error or negative maximal error derivative; “4” corresponds to zero error or zero error derivative; “8” corresponds to positive maximal error or positive maximal error derivative, respectively. According to e(t) and de(t)/dt, the corresponding fuzzy logic control actions u(t) can be generated, which is also divided into nine parts, as “0” denotes zero control output and “8” denotes the maximal control output. For example, as indicated in Table I, negative maximal (0) e(t) and negative maximal (0) de(t)/dt will generate minimal (0) u(t), and positive maximal (8) e(t) and positive maximal (8) de(t)/dt will generate maximal (8) u(t). The corresponding membership functions of e(t), de(t)/dt and u(t) are given in Fig. 5. Each of them adopts triangle form and is decomposed in nine sections. The nine subsections of the membership functions marked with 0 ~ 8 correspond with the nine subsections 0 ~ 8 in Table I, respectively. The peak of each triangle has certainty ratio “1”, while the bottom of each triangle has certainty ratio “0”. The centre value of each triangle for e(t), de(t)/dt and u(t) are also given in Fig. 5. For example, the triangle centre value of e(t) is specified as -4A, -3A, -2A, -1A, 0A, +1A, +2A, +3A, +4A, respectively. In order to regulate the scaling, e(t) and de(t)/dt have their respective scaling gains: ke and kse. Studies show that fuzzy controller with constant ke have poor control performance considering different operating points, such as large steady-state error with small reference resonant effective current iL_eff*, or obvious ripple for large reference effective current iL_eff*. In order to solve this problem, an adaptive gain ke is developed according to iL_eff*. Since u(t) is an intermediate fuzzified controller output, it should be further transferred to real control variables: fs(t) and d(t). According to Eq.(1), Eq.(2) and Table I, it can be concluded that d(t) has a positive proportional relationship with u(t), while fs(t) has a negative proportional relationship with u(t). The corresponding transition diagram between u(t) and fs(t), d(t) can be obtained, as shown in Fig. 6. 4 Table I Rule base of fuzzy PD controller 0 1 2 error e(t) 3 4 5 6 7 8 0 0 0 0 0 1 2 2 3 8 1 0 0 1 1 2 2 3 4 8 2 0 1 2 2 2 3 4 5 8 3 4 5 6 7 0 0 0 0 0 1 2 2 3 4 2 2 3 4 5 2 3 4 5 6 3 4 5 6 6 4 5 6 6 7 5 6 6 6 7 6 6 7 7 8 8 8 8 8 8 8 0 5 6 6 7 8 8 8 8 output u(t) error derivative de(t)/dt Fig. 5 Membership functions of inputs and output of the fuzzy PD controller The resulting current-mode fuzzy PD control diagram is depicted in Fig. 6: it comprises a cascaded two-closed-loop control for the LCC resonant converter. The outer is a conventional linear control, which regards the output voltage uo as the control object. This content will not be discussed here. The inner is the current-mode fuzzy control loop, which regards the resonant effective current iL_eff as the control object, which is the emphasis of this paper. As shown in Fig. 6, the reference current iL_eff* compares with the measured current iL_eff, the difference between them is the error signal e(t), which is further deduced for its derivative de(t)/dt. After regulation by adaptive scaling gain ke and constant scaling gain kse, the error signal e(t) and derivative signal de(t)/dt are fed into the fuzzy controller. According to the membership functions, the fuzzification procedure is executed and both inputs in real-valued space are transferred to a fuzzy space. Next, referring to the rule base, the inference mechanism is activated and the reasonable control actions are derived. Adopting the “centre of gravity (COG)” defuzzification method, the fuzzy controller output u(t) is generated from a fuzzy space to a real-valued space, which is then further transferred to actual control variables: fs(t) and d(t). As denoted in Fig. 6, with a specified modulation limit, fs(t) is negative proportional to u(t) with a positive 5 Fig. 6 Current mode fuzzy control block diagram of LCC resonant converter offset, while d(t) is positive proportional to u(t) without offset. The generated fs and d are provided to a pulse width modulator, the resulting transistor gate drive signals are given to transistors of the fullbridge inverter. The resonant effective current iL_eff is measured and feedback via appropriate filter and compared with the reference signal iL_eff* to complete the current-mode fuzzy closed-loop. IV. Simulation Studies In order to ensure the correctness and effectiveness of the current-mode fuzzy PD controller, some simulation and comparison are executed. As shown in Fig. 7, the blue curve is the reference sinusoidal resonant effective current iL_eff* with magnitude from 1 A to 19 A with increasing frequency from 10 Hz to 100 Hz in 0.14 s, while the red curve is the measured resonant effective current iL_eff from LCC resonant converter in MATLAB. From Fig. 7, it can be concluded that the plant resonant current agrees well with the reference current in the complete magnitude range and frequency range besides a little spikes at peak reference value. Fig. 8 shows the simulation results with a square form reference resonant current. The blue is the referenced 50 Hz pulsing iL_eff* with magnitude from 4 A to 19 A, while the red is the measured simulated resonant current iL_eff with the developed fuzzy PD controller. It can be seen that the red follows the step-change blue in about 2 ms without overshoot. The exclusive shortage is that there exists some vibration (about 5%) around the steady-state large reference current. But since a cascaded two-closed-loop is adopted for the LCC resonant converter, such vibrated signal can be attenuated through the additional outer voltage controller. Such estimation should be verified through later study and experiment. In order to show the advantages of the fuzzy PD controller, a comparison between the fuzzy controller and a linear PI controller is implemented and the results are also shown in Fig. 8. Among which, the light blue is the measured resonant effective current iL_eff with a conventional PI controller. From the comparison, it can be seen that the plant resonant effective current iL_eff with the fuzzy controller has smaller ripple than that with a linear PI controller, which demonstrates the benefits of the novel fuzzy control approach. 6 Fig. 7 Referenced sinusoidal increasing frequency current and plant response with fuzzy controller Fig. 8 Referenced pulse current and plant response with fuzzy and linear controller V. Conclusions A novel current-mode fuzzy PD controller with adaptive gain is developed for the LCC resonant converter applied in very low frequency high voltage generators using switching frequency and duty cycle control variables. Through simulation studies, the developed fuzzy control shows satisfying performance for various operating points. The comparison between the fuzzy PD control and a conventional linear PI control shows performance improvement with smaller ripple of the novel fuzzy control approach. The advantage of the developed fuzzy controller is not so dominant when compared with a standard linear controller. In the other aspect, such model-free fuzzy control always suffers from criticism of lacking of systematic design with consistent and guaranteed performance. Due to such shortage, a model-based fuzzy control with systematic stability analysis and controller design is expected in the future. 7 References [1] VLF related Standards: VDE DIN 0276-620, IEEE P400.2, VDE DIN 0276-621, CENELEC HD 620 and CENELEC HD 621. [2] Juan A. Martin-Ramos, Juan Diaz, Alberto M.Pernia, Juan Manuel Lopera, Fernando Nuno, “Dynamic and steady-State Models for the PRC-LCC Resonant Topology With a Capacitor as Output Filter,” IEEE Transactions on Industrial Electronics, Vol.54, No.4, August 2007. [3] M. Hu, N. Fröhleke and J. Böcker, “Small-Signal Model and Control Design of LCC Resonant Converter with a Capacitive Load Applied in Very Low Frequency High Voltage Test System,” IEEE Energy Conversion Congress and Expo (ECCE), 2009. [4] Eric X. Yang, Byungcho Choi, Fred C.Lee, and Bo H.Cho, “Dynamic Analysis and Control Design of LCC Resonant Converter,” Power Electronics Specialists Conference, PESC’92, 23rd Annual IEEE, Vol.1, pp.362-369, 1992. [5] Heiko Osterbolz, Cornelius Paul, Philips Medical Systems, “Study of Resonant Hgh-Voltage Cascaded Circuits with Different Numbers of Stages,” Simulation in Drive Technology, Power Electronics and Automotive Engineering. SIMPLORER Workshop 2001. [6] Z. Cao, M. Hu, N. Froehleke, J. Boecker, “Modeling and Control Design for a Very Low-Frequency HighVoltage Test System,” IEEE Transactions on Power Electronics 25(2), 1068-1077, 2010. [7] Gang Feng, “A Survey on Analysis and Design of Model-Based Fuzzy Control Systems,” IEEE Transactions on Fuzzy systems, Vol.14, No. 5, Oct. 2006. [8] Piero P. Bonissone, Pratap S. Khedkar, Michael J. Schutten, “Fuzzy Logic Control of Resonant Converters for Power Supplies,” Control Applications, 1995, Proceedings of the 4th IEEE Conference. [9] R.B.Ridley, B.H.Cho, and F.C.Lee,”Analysis and interpretation of loop gains of multi-loop-controlled switching regulator,” IEEE Trans. Power Electron, pp.489-497,1988. [10] Han-Xiong Li, Lei Zhang, Kai-Yuan Cai, GuanrongChen, “An improved Robust Fuzzy-PID Controller With Optimal Fuzzy Reasoning,” IEEE Transactions on Systems, Man and Cybernetics-Part B: Cybernetics, Vol. 35, No. 6, December 2005. [11] George K. I. Mann, Bao-Gang Hu, Raymond G. Gosine, “Analysis of Direct Action Fuzzy PID Controller Structures,” IEEE Transactions on Systems, Man and Cybernetics-Part B: Cybernetics, Vol. 29, No. 3, June 1999. 8