High-Performance DC-AC Power Converter Using Adaptive Control Technique for Advanced Material Machining
Total Page:16
File Type:pdf, Size:1020Kb
Journal of Automation and Control Engineering Vol. 5, No. 1, June 2017 High-Performance DC-AC Power Converter Using Adaptive Control Technique for Advanced Material Machining En-Chih Chang Department of Electrical Engineering, I-Shou University, No.1, Sec. 1, Syuecheng Rd., Dashu District, Kaohsiung City 84001, Taiwan, R.O.C. Email: [email protected] Wei Liu and Hongshu Huang Electrical Engineering and Automation, Guangzhou University, Guangzhou 510006, P.R.C. Email: [email protected], [email protected] Rong-Ching Wu Department of Electrical Engineering, I-Shou University, No.1, Sec. 1, Syuecheng Rd., Dashu District, Kaohsiung City 84001, Taiwan, R.O.C. Email: [email protected] Abstract—A high-performance DC-AC power converter possible. To obtain these requirements, proportional- using adaptive control technique for advanced material integral (PI) controller is one of the useful ways to machining (AMM) is developed in this paper. The presented achieve suitable characteristics. But, PI is difficult for technique combines the merits of hyperbolic tangential converter control to gain good steady-state and dynamic sliding surface (HTSS) and adaptive neuro-fuzzy inference response under nonlinear loads [5-8]. Many control system (ANFIS). The HTSS not only has the robustness of classic SMC but increases the system trajectory convergence techniques are also used in literature, such as mu- speed. However, once a highly nonlinear load occurs, the synthesis, H-infinity control, wavelet control, and so on chattering still exists. The chattering causes high total [9-14]. However, these techniques are complex, and harmonic distortion (THD) in DC-AC power converter difficult to realize. Proposed in 1950s, sliding mode output voltage, thus yielding the instability and unreliability control (SMC) provides insensitivity to parameter of advanced material machining. The ANFIS is thus variations and removal of disturbances [15], [16]; SMC employed to eliminate the chattering and simultaneously the of DC-AC power converter systems are also developed. HTSS provides finite system-state convergence time. By However, these SMC systems use linear sliding surface, combining HTSS with ANFIS, the DC-AC power converter and such surface has infinite system-state convergence of AMM will achieve robust performance. Experimental results are given to conform that the proposed technique time [17-20]. Thus to increase the convergence speed, a can obtain low total harmonic distortion (THD) and fast hyperbolic tangential sliding surface (HTSS) can be used. transience even under phase-controlled loads. Owing to the Compared with linear sliding-surface-based, the HTSS notable superiorities, e.g. computational quickness, practical can enforce system tracking error to converge to zero in simplicity, and programmable easiness in the proposed finite time [21-24]. However, when a severe nonlinear technique, this paper will be a helpful reference to load is applied, the HTSS system still exists in chattering researchers of related advanced material machining. problem. The chattering will cause high THD in DC-AC power converter output and the stability and reliability of Index Terms—DC-AC power converter, advanced material the AMM are thus deteriorated. Adaptive neuro-fuzzy machining (AMM), hyperbolic tangential sliding surface (HTSS), adaptive neuro-fuzzy inference system (ANFIS) inference system (ANFIS) associating the training ability of neural network with approximate human reasoning of fuzzy logic is well-known artificial intelligence approach. I. INTRODUCTION By such a hybrid learning approach, the error between The DC-AC power converters are popularly used in desired and actual output can be minimized [25-28]. advanced material machining (AMM), such as nickel- Therefore, the ANFIS is used to eliminate the chattering based super alloys, nuclear materials, and titanium alloys while the HTSS is utilized to achieve finite system-state [1-4]. The requirements of a high-performance DC-AC convergence time. When this proposed technique is power converter include output voltage with low THD applied to the control of a DC-AC power converter of even under nonlinear loads, fast transient under phase- AMM, the system will show the effectiveness of low controlled load and steady-state errors as small as output voltage distortion, fast transient and insensitivity to load disturbances. Experimental results demonstrate Manuscript received October 28, 2016; revised July 19, 2017. ©2017 Journal of Automation and Control Engineering 5 doi: 10.18178/joace.5.1.5-9 Journal of Automation and Control Engineering Vol. 5, No. 1, June 2017 the feasibility and advantages of using the proposed (4) e e 2 2 2 technique. cosh (e1) II. PROPOSED ADAPTIVE CONTROL TECHNIQUE FOR The equivalent control, can be formulated as DC-AC POWER CONVERTER OF AMM 0 (5) f 0, u u The system block diagram of DC-AC power converter p e of AMM is illustrated in Fig. 1. Define v be the output o The us is used to suppress the system uncertainties, voltage, vref be the desired sinusoidal waveform, and and can be obtained via the following inequality. ve vo vref be the voltage error. Suppose that the 0 (6) u u u switching frequency is high enough, DC-AC power P e s converter can be considered as a constant gain, K pwm . Then, substituting (1) and (5) into (6), we have Let e1 ve and e1 e2 , the error dynamic matrix can be us K sgn(e1), K 0 (7) written as Though the (7) provides finite system-state e 0 1 e 0 0 convergence time to the origin, the chattering will occur 1 1 K 1 1 pwm u (1) when a highly nonlinear load is applied. Thus, one e e f 2 LC RC 2 LC solution eliminates the chattering via ANFIS as follows. Firstly, a Takagi-Sugeno (T-S) fuzzy formation is 1 1 established by the ANFIS and then we model the given where f vref vref vref is the system LC RC training data. Thus, the ANFIS can be expressed as uncertainty. Rule i : If e is A and e is A Full- Bridge Inverter Filter 1 i1 n in (8) Then u p e p e r L i i1 1 in n i iL io + T1 T3 where Rule i represents the ith fuzzy rules, i 1, 2, j , ic Aik is the fuzzy set in the antecedent associated with the Vdc C R kth input variable at the fuzzy rule, and pi1,pin, T2 T4 ri symbol the fuzzy resulting parameters. - Using the ANFIS, the yields up w1u1 wju j (9) PWM Generator vo where w w w w and w w w w . u + 1 1 1 j j j 1 j Gate p e1 Proposed Technique - v Driver ref Due to u p e p e r , the (9) can be Circuit Comparator i i1 1 in n i Triangualr Wave rewritten as DSP - Based Controller u p w1u1 w ju j Figure 1. DC-AC power converter of AMM. (w1e1) p11 (w1en) p1n w1r1 (10) The control objective is to design a control law u p well so that the output voltage can equal the desired (w je1) p j1 (w jen) p jn w jrj reference sinusoidal. The is designed below. The (10) implies that the ANFIS with the formation of five-layer and such formation can be displayed as Fig. 2. up ue us (2) Layer I Layer II Layer III Layer IV Layer V where the equivalent control, ue is valid only on the A11 sliding surface, and the sliding control, u compensates w1 w1 s e1 N w1u1 for the disturbance effects, and then the system will reach A21 the sliding surface and converge in finite time. To accelerate the convergence speed of the sliding u p phase and reaching phase, a hyperbolic tangential sliding A12 surface is created as w2 w2 e2 w2u2 e tanh(e ) (3) 2 1 A22 where and are constants. Figure 2. ANFIS formation From (3), we have ©2017 Journal of Automation and Control Engineering 6 Journal of Automation and Control Engineering Vol. 5, No. 1, June 2017 III. EXPERIMENTAL RESULTS The experimental verification of the proposed technique is carried out on a DC-AC power converter with specifications shown in Table I. TABLE I. SYSTEM PARAMETERS Filter inductor L=0.5 mH Filter capacitor C=10 μF Figure 5. Experimental waveforms for the proposed technique with DC supply voltage Vdc=200 V phase-controlled load (100V/div; 20A/div; 5ms/div) Resistive load R=10 Ω Output voltage and frequency vo=110 Vrms, f=60 Hz Switching frequency fs =18 kHz We presume that the LC filter parameter values are varied from 10% ~ 200% of nominal values as the proposed system is under 10Ω resistive load; Fig. 3 and Fig. 4 show experimental output voltage waveforms with the proposed technique and the classic SMC, respectively. Owing to the elimination of the chattering, the proposed technique is insensitive to LC parameter variations than Figure 6. Experimental waveforms for the classic SMC with phase- controlled load (100V/div; 20A/div; 5ms/div) the classic SMC. To examine the transient behaviour of the DC-AC power converter of AMM, Fig. 5 shows L i experimental output voltage and the load current for the L T1 T3 proposed technique with phase-controlled load. As can be io ic seen, a fast recovery of the steady-state response and + Vdc slight voltage depression can be obtained; Chowever, the - Load classic sliding mode controlled DC-AC power converter T2 T4 of AMM, shown in Fig. 6 indicates large voltage depression and slow recovery time. Therefore, the proposed technique furnishes higher tracking precision, faster convergence, better voltage compensation and greater robustness. The block diagram of the hardware Gate driver circuit implementation is represented as Fig. 7. The output v voltage %THD in the face of LC variations and phase- Dead time circuit O Voltage controlled load are given in Table II.