Article Preview
Top1. Introduction
In recent years significant advance have been made on the sensorless field oriented controlled induction machines drives. In the middle of 1980 direct torque control was developed by Takahashi and Depenbrock (Bose, 2002) as an alternative to field oriented control to overcome its problems (Sorchini & Krein, 2006) many techniques are used for implementation of DTC as in Nikolic and Jeftenic (2008) and Casadei et al. (2006).The DTC model used for simulation is illustrated using MATLAB simulink for linear model of induction motor in Wahab and Sanusi (2008). In a direct torque controlled (DTC) induction motor drive supplied by a voltage source inverter, the scheme as the name indicates, it is possible to control directly the stator flux linkage (or rotor flux or the magnetization flux) and the electromagnetic torque by the selection of an optimum inverter voltage vector through a pre designed inverter gate pulses look-up tables. The stator flux and electromagnetic torque are controlled in a closed loop by using flux and torque hysteresis comparators (Mohanty, 2009). Then, the selection of the optimum voltage vector of a voltage source inverter is to restrict the flux and torque errors within their respective flux and torque hysteresis comparators bands, to obtain fast torque response, low inverter switching frequency and low harmonic losses (Purcell & Acarnley, 2001; Haddoun & Benbouzid, 2007). Unlike field oriented control method, DTC techniques require utilization of hysteresis band comparators instead of flux and torque controllers In addition, utilization of look-up tables to select inverter switching gate pulses is to replace the coordinate transformation and pulse width modulation (PWM) signal generators of field oriented control (Casadei & Tani, 2002; Lee & Blaabjerg, 2006). Many techniques are used for the design of speed and flux observers as in Lee et al. (2005) and Salvatore and Stasi (2000) that used for estimation of the motor speed and fluxes without a need of speed and flux sensors that can reduce the cost are. Many techniques are developed to improve the performance of the DTC, such that estimation of the motor parameters deviation from its nominal values during the operation as in Capolino and M (1990) and Jarray and Tholomier (2000). Artificial intelligent techniques are also used to improve the performance of DTC as in Meziane and Benalla (2006) and Babu and Das (2010). The common assumption made in the development of these control laws is the linearity of the magnetic circuit of the machine. This assumption is usually justified by including the flux magnitude in the outputs to be regulated by the controller and keeping this magnitude regulated at a value far from the saturation region (Marino et al., 1993; Casadei et al., 2004). However there are no guarantees that the flux magnitude remains in the linear magnetic region during machine transients. Moreover in many variable torque applications, it is desirable to operate the machine in the magnetic saturation region to allow the machine to develop higher torque (Sullivan & Sanders, 1995; Hofmann et al., 1997). Saturation effects are also known to be pronounced in drives operating in the field weakening region, or in drives that operate with varying flux levels to achieve optimally in a specified sense (Kirshnen et al., 1985). However, the operation of the motor at various magnetization levels makes the nominal inductance a bad approximation. Recently, researchers have been attracted to induction motor control with magnetic saturation. Feedback input-output linearization schemes for induction motors with magnetic saturation were proposed in a fixed stator frame (Tarbouchi & LeHuy, 1998) and in a synchronously rotating frame (Novotnak et al., 1999). While in Tarbouchi and LeHuy (1998) the control signal is the stator voltage, in Novotnak et al. (1999) it is the stator current. Both papers treat the T-model of an induction motor. Unfortunately, due to the complicated nature of the T-model, drastic simplifications are required to facilitate the use of this model in nonlinear control synthesis. The major drawback in Tarbouchi and LeHuy (1998) (also present in the optimal flux reference selection of Novotnak et al., 1999) is the assumption that the stator and rotor leakage parameters ¾s and ¾r which are the scalar saturation functions, as defined in Leonard (1985), are equal and constant. This assumption has the indirect effect of neglecting any cross-saturation effect that might appear in the dynamics of the motor. On the other hand, the model in Novotnak et al. (1999) is obtained by firstly simplifying the motor equations assuming a linear magnetic circuit and then including a mutual inductance that varies with mutual current. This approach does not include derivatives of the saturation function that should appear in a complete model (Levi, 1995). A similar modeling approach can also be found in Gokdere (1996) for incorporating magnetic saturation in the passivity- based control design methodology of Nicklasson (1997). It is worth pointing out that, in Gokdere (1996) similar to Novotnak et al. (1999), stator currents are used as the control signal. All the work presented so far is based on a T -model of the induction motor, contrary to the π-model proposed in Sullivan and Sanders (1995). The π-model differs from the conventional T-model in that it is more closely related to the physical structure of the machine, since its derivation is primarily based on the stator-rotor tooth pair magnetic circuit. Even though the work in Sullivan and Sanders (1995) is based on a wound rotor motor, it is shown in the same paper how the modeling approach can be applied to a squirrel cage motor. It is not difficult to show that both models are equivalent when a linear magnetic circuit is assumed, this equivalence does not hold when main flux saturation is included. In the published work Fattah and Loparo (1999), it was shown that considering magnetic saturation explicitly in nonlinear control synthesis is of foremost importance especially when the machine is voltage actuated. Because the π- model was experimentally found in Sullivan and Sanders (1995) to be better suited to capture the nonlinear magnetic effects. In this paper, the induction motor with magnetic saturation is considered with the DTC model, that no simplifying assumptions are used in the development of the model. This paper is different from the previous techniques of DTC speed sensorless systems that we are using the saturated model of induction motor instead of simplicity assumption of linear magnetic motor model with Lyapunov and ANFIS speed observers. Also this research is study how to estimate the variation in the motor stator resistance during the operation by using ANFIS and fuzzy techniques.