TMCnet News
Low Pass Filtering Based Artificial Neural Network Stator Flux Estimator for AC Induction Motors [Sensors & Transducers (Canada)](Sensors & Transducers (Canada) Via Acquire Media NewsEdge) Abstract: A novel artificial neutral network stator flux estimator based on low pass filter for induction motors is proposed in this paper. Firstly, low pass filter is used to extract the features of the stator voltage. Then, the feature signals are selected as inputs of an artificial neutral network regression model to estimate stator flux. The combination stator flux estimation algorithm with low pass filter and artificial neutral network reduces the interference induced by high frequency signals and has fine dynamic performance. Simulations show high performance of the proposed stator flux estimator under different torques. Copyright © 2013 IFSA. Keywords: Induction motor, Direct torque control, Stator flux estimation, Artificial neutral network, Low pass filter. (ProQuest: ... denotes formulae omitted.) 1. Introduction Owing to their simple structure and stability, AC induction motors (IM) have been widely used in many fields [1]. An accurate flux estimation is very important to a high performance induction motor drive, such as Direct Torque Control (DTC) [2], which has attracted more and more attention in recent years, due to its simple structure and ability to achieve fast torque and flux control. A typical IM has no flux sensors built in, and there is no place provided in it to mount such sensors. Flux can be obtained by observers based on mathematic models [3, 4] such as Kalman filters [5], Luenberger observers [6-8], etc. Most of the flux estimators are based on the voltage model, the current model, or a combination of both [3, 10]. The estimator based on the current model requires the knowledge of stator current and rotor speed. In some industrial applications, the use of an incremental encoder to get the speed or position of the rotor is undesirable since it reduces the robustness and reliability of the drive. It has been widely known that even though the current model has managed to eliminate the sensitivity to the stator resistance variation. The estimators are sensitive to the rotor parameter, especially at high rotor speed [3, 10, 11]. The voltage model is normally used at high speed. And some problems arise at low speed [9-12]. In practice, even a small DC offset in the back electromotive force (EMF) can cause the integrator to saturate [12, 13]. To overcome this, a low pass filter (LPF) is normally used in place of a pure integrator [13, 14]. However, compared to the pure integrator, particularly at frequencies close to the cut-off, LPF will result in phase and magnitude errors. Attempts have been made to improve the estimated stator flux based on the LPF as given by [13-16]. In [13], the proposed method creates a novel arithmetic between the pure integrator and the LPF. The method uses an adaptive control system, which is based on the fact that the back EMF is orthogonal to the stator flux. The compensator is adapted for this condition. However, implementation of the proposed system requires large processor resources and significantly increased the complexity of the control system. In [14], the errors between LPF and pure integrator were computed and compensated. But the compensation is based on the steady-state condition and only can be used under steady-state condition. In [15], two different cut-off frequencies were used in back EMF equation. But it will degrade the ability of driving to some extent. A new type of stator flux integration has been proposed in [16]. Output of the flux is used for feedback based on first-order LPF. Control effect of the algorithm is remarkable. But it is more complex. By analyzing the error of the LPF based estimator, a simple compensation method has been proposed [17]. This method is mainly applied to solve the steady-state situation problem. And it is largely ineffective to improve the flux estimation accuracy at low speed. The voltage model has been used in the high speed area and current model in the low speed area [18]. The disadvantage of this hybrid model is that it is very difficult to switch quickly and smoothly between the two models. In this paper, to overcome the problem of the integration for stator flux estimation, a novel artificial neutral network stator flux estimate method based on low pass filter is proposed. The proposed method is very simple. And it is a Wiener model in nature, which consist of a linear dynamic (LPF) and a nonlinear static (ANN), yet it can improve the dynamic performance of stator flux estimation. Simulations show the high performance of the method. 2. Stator Flux Estimation 2.1. Principle of Flux Estimation Based on LPF The stator flux calculation based on the stator voltage equation is given by [19]: ... (l) Its frequency expression is: ... (2) where l//s is the stator flux, Rs is the stator resistance, and us and is are the measured terminal voltages and currents, respectively. COe is the motor stator signals when steady-state operation of angular. The integration of (1) by pure integrator suffers the drift and the saturation problems. To solve the problems, traditional strategy is to replace the pure integrator by a LPF. The structure of the stator flux estimation system is showed in Fig. 1. Then the Equation (1) can be written as ...(3) where COc is the cut-off frequency of the LPF. By comparing (2) with (3), ...(4) Actually this method will degrade the performance of the system. From the above equations, the magnitude of the estimated stator flux is always less than the actual one which can result in magnetic flux saturation in low speed region. On the other hand the phase error will also lead to incorrect voltage vector selection. When the estimated flux enters a new sector, the actual flux is still in the previous sector, so the voltage vector will be selected incorrectly. 2.2. Cascaded LPF Based Flux Estimation A programmable cascaded LPF is proposed to solve the drift problem and to estimate exactly stator flux [9, 20]. The principle of the cascaded LPF method of integration can be explained as follows. Since the drive has to operate in a wide frequency range, a single-stage integrator has to be designed with a very large time constant. This causes the problem of DC offset and its very slow decay, as dictated by the time constant. If a single-stage integrator is resolved into a number of cascaded LPF with a short time constant, the problem of DC offset decay time can be sharply attenuated [9]. The structure of the stator flux estimation system is showed in Fig. 2. The programmable cascaded LPF perform back EMF integration. The algorithm described in [9] does not introduce acceptable estimation when a direction of a stator field rotation is changing. And the scheme also has a drawback in that the time constant of the LPF will be very large at times [9]. 2.3. Artificial Neural Network Based Flux Estimation Artificial neural networks (ANNs) are suitable for AC motor state estimation, because of their known advantages, such as the ability to approximate any nonlinear functions to a desired degree of accuracy, learning and generalization, fast parallel computation, robustness to input harmonic ripples, and fault tolerance [21, 22]. These aspects are important in the case of nonlinear systems, like converter-fed AC drives, where linear control theory cannot be directly applied. Additionally, highefficiency power electronic converters used for ac motors operate in switch mode, which results in very noisy signals. For these reasons, ANNs are attractive for signal processing and control of AC drives. The usual ANN model is the multilayer feedforward network using the error back propagation algorithm (BP) [23]. The artificial neural network can be used directly to design a new observer of the stator flux. The idea is to model flux directly using stator currents and voltage as inputs instead of back EMF. The structure of the ANN based stator flux estimation system is showed in Fig. 3. Despite many advantages, the ANN estimator has serious limitations inherited. It requires that initially selected sampling time is applied to the data in learning. If we want the estimated flux to be smooth, short sampling time should be considered, which can result in large training set. And the ANN model is a static model without fine dynamic performance. 2.4. LPF Based Artificial Neural Network Stator Flux Estimation Inspired by the above methods, we present the idea of combining LPF and ANN as a new stator flux estimator. Fig. 4 shows the LPF based ANN stator flux estimator structure. The cut-off frequency coc of the LPF influences the errors of the flux estimator greatly. In practice, the cut-off frequency can't be set too small since the stability of the control system will be degraded when the cut-off frequency is too small. According to [24], we set the coc at 30 rad/s. 3. Simulations and Experiment results In order to get the data, a converter-fed DTC system is carried out. The training patterns are prepared by numerical simulations of the induction motor model in the stationary reference frames (a-ß) with help of MATLAB and SIMULINK. In simulations the nominal data of 3.7 kW induction motor is used. Table 1 shows the parameters of induction motor. The design and training of a neural network for satisfactory performance requires very time consuming iterative procedure with large training data table. At last we chose the data in the condition of the motor at 1000 rpm and under the rated load torque TL 25 Nm. By less than 600 times training, the ultimate error is less than 1.00e-5. The presented results were obtained by the LPF based ANN estimator in steady-state under the TL at 1000 r/m in Fig. 5. From Fig. 5, it can be concluded that the new estimator shows high performance. Fig. 6 to Fig. 12 show the robustness performance of the proposed method under different load torques. Fig. 10 to Fig. 12 show the results when load torque exceeds the training range (from (1.2-1.6) Tl). 6. Conclusions A novel LPF based ANN stator flux estimator is proposed in the paper. Simulations show the LPF based ANN estimator perform good performance of robustness to torque changes. This method has some certain prospect in the estimation of stator flux of AC induction motors. Acknowledgements This work is supported by the Priority Academic Program Development of Jiangsu Higher Education Institutions under Grant PAPD [2011] 6, China Postdoctoral Science Foundation under Grant 20110491359, Jiangsu Postdoctoral Sustentation Fund, China under Grant 1102109C, and the open fund of Key Laboratory of Measurement and Control of Complex Systems of Engineering, Ministry of Education under Grant MCCSE2012A03. References [1] . V. N. Walivadekar, H. Manjunath, V. S. Babu and Sadda K., Development of transistorised PWM inverter fed induction motor drive for electric vehicle-a case study, in Proceedings of the International Conference on Power Electronics, February 1996, pp. 741-747. [2] . I. Takahashi, T. Nogushi, A new quick-response and high-efficiency control strategy of an induction motor, IEEE Trans. Ind. Appl., 1A, 22, October 1986, pp. 820-827. [3] . P. L. Jansen and R. D. Lorenz, A physically insightful approach to the design and accuracy assessment of flux observers for field oriented induction machine drives, IEEE Trans. Ind. Applicat., 30, January/February 1994, pp. 101-110. [4] . P. Vas, Sensorless vector and direct torque control, ser. Monographs in electrical and electronic engineering, 42, Oxford University Press, 1998. [5] . Texas Instruments Europe, Sensorless control with Kalman filter on TMS320 fixed-point DSP, Literathre number: BPRA057, 1997. [6] . H. Kubota and K. Matsuse, Speed sensorless fieldoriented control of induction motor with rotor resistance adaptation, IEEE Trans, on Ind. App., 30, 5, September/October 1994. [7] . H. Kubota, K. Matsuse and T. Nakano, DSP-based speed adaptive flux observer of induction motor, IEEE Trans, on Ind. App., 29, 2, March/April 1993. [8] . J. Maes and J. A. Melkebeek, Speed-sensorless direct torque control of induction motors using an adaptive flux observer, IEEE Trans, on Ind. App., 36, 3, May/June 2000. [9] . B. K. Bose and N. R. Patel, A programmable cascaded low-pass filter-based flux synthesis for a stator flux-oriented vector-controlled induction motor drive, IEEE Trans. Ind. Electron., 44, 1, February 1997, pp.140-143. [10] . M. Elbuluk, N. Langovsky and M. D. Kankam, Design and implementation of a closed-loop observer and adaptive controller for induction motor drives, IEEE Trans. Ind. Applicat., 34, May/June 1998, pp. 435-443. [11] . T. G. Habetler, F. Profumo, G. Griva, M. Pastorelli, A. Bettini, Stator resistance tuning in stator flux field-oriented drive using an instantaneous hybrid flux estimator, IEEE Trans. Power Electron., 13, January 1998, pp. 125-133. [12] . K. D. Husrt, T. G. Habetler, G. Griva and F. Profumo, Zero speed tacholess IM torque control: Simply a matter of stator voltage integration, IEEE Trans. Ind. Appl., 34, July/August 1998, pp. 790-795. [13] . Jun Hu, Bin Wu, New integration algorithms far estimating motor flux over a wide speed range, IEEE Transactions on Power Electronics, 9, 1998, pp. 969-977. [14] . N. R. N. Idris and A. H. M. Yatim, An improved stator flux estimation in steady-state operation for direct torque control of induction machines, IEEE Trans. Ind. Appl., 38, 1,2002, pp.l 10 -116. [15] . Jiaiong Wu, Yongdong Li, Jie Chen, Hu Hu, Speed sensor-less direct torque control of an induction machine in low speed region, PEAK, 2000, pp. 461-468. [16] . Mustapha Messaoudl, A Robust Sensorless Direct Torque Control of Induction Motor Based on MRAS and Extended Kalman Filter, Leonardo Journal of Sciences, 12, January 2008, pp. 35-56. [17] . R. Krishnan and A. S. Bharadwaj, A review of parameter sensitivity and adaptation in indirect vector controlled induction motor drive systems, IEEE Trans. J. Power Electron, 6, 4, February 2007, pp. 623-635. [18] . N. Benaidja, Identification of Asynchronous Machine Parameters by Evolutionary Techniques, Electric Power Components and Systems, 34, 12, January 2006, pp. 1359-1376. [19] . X. Xu and D. W. Novotny, Implementation of direct stator flux oriented control on a versatile DSP-based system, IEEE Trans. Ind. Applicat, 27, July/August 1991, pp. 694-700. [20] . B. K. Bose, M. G. Simoes, D. R. Crecelius, K. Rajashekara and R. Martin, Speed sensorless hybrid vector controlled induction motor drive, in Proceedings of the Conf. Rec. IEEE-IAS Annu. Mtg., 1995, pp. 137-143. [21] . M. P. Kazmierkowski, R. Krishnan and F. Blaabjerg, Control in Power Electronics Selected Problems, Academic, New York, 2002. [22] . B. K. Bose, Modem Power Electronics and AC Drives, Pearson Education, Prentice-Hall, Englewood Cliffs, NJ, 2002. [23] . Wang Yongji and Tu Jim, Artificial Neural Network Control, China Machinery Industry Information Institute, Beijing, 1998. [24] . Chen Chang, Huang Xiayan, Method of stator flux estimations with variable cutoff frequency, Electronic Instrumentation Customer, 2011. 1,2 Yuhan Ding,1 Shaoqing Zhou,1 Congli Mei,1 Hui Jiang 'JiangsuUniversity, Zhenjiang, 212013, China 2 Key Laboratory of Measurement and Control of CSE (School of Automation, Southeast University), Ministry of Education, Nanjing, China E-mail: [email protected], [email protected] Received: 18 September 2013 /Accepted: 22 November 2013 /Published: 30 December 2013 (c) 2013 International Frequency Sensor Association |