Minimum Motor Losses Point Tracking for a Stand-Alone Photovoltaic Pumping System Tomás Corrêa∗ , Francisco A. S. Neves†, Seleme I. Seleme Jr.∗ and Selênio R. Silva∗ ∗ PPGEE/UFMG - Belo Horizonte, MG, Brazil E-mail: tpcorrea@uol.com.br, seleme@ppgee.ufmg.br and selenios@ppgee.ufmg.br † DEESP/UFPE - Recife, PE, Brazil E-mail: fneves@ufpe.br Abstract— Solar photovoltaic energy has been, for many years now, a promising alternative in the diversification of global energetic sources. Although photovoltaic energy has many advantages, it has not yet become commercially popular, because it is still an expensive alternative. In fact, studies have already shown that the investment in such systems are paid back, but it is still hard to find people willing to wait five or six years to have their money back. The aim of this work is to reduce the payback time of photovoltaic stand-alone pumping systems, optimizing not only the efficiency of the photovoltaic conversion, using a maximum power point tracking algorithm, but also minimizing the losses in the induction motor. Based on the design procedures described in this paper, a prototype has been developed and experimental results with the optimizing strategie are shown. I. I NTRODUCTION The process of photovoltaic energy conversion has special intrinsic features which, for a given condition on solar irradiance and panel temperature, make possile to find an operation point where the conversion efficiency is maximum. Fig. 1 shows typical curves of a photovoltaic array (PV) under certain conditions on the temperature (curves a and b) and irradiance (curves c and d), where curves a and b show the array current (Ipv ) as function of voltage (Vpv ) and curves b and d show the array power (Ppv ) as function of voltage. (a) Current [A] 4 25oC 55oC 3 (b) 1500 2 Power [W] 5 25oC 1000 55oC 500 1 0 0 200 300 400 Power [W] 3 2 1 200 Wm−2 0 100 200 300 400 Voltage [V] Fig. 1. 0 100 500 200 300 400 500 (d) 1500 1000 Wm−2 4 0 0 500 (c) 5 Current [A] 100 1000 Wm−2 1000 500 0 200 Wm−2 0 100 200 300 400 Voltage [V] 500 Typical curves of a PV array. Authors would like to thank Eletrobrás/LENHS and CAPES for the financial support. Considering that the prices of solar panels are still quite high and the variation of their efficiency with the operating point, as seen from Fig. 1, the use of some technique for tracking the point of maximum power (MPPT) is imperative. This has been a subject of interest of several studies and many approaches have been proposed in the literature [1]. Among the existing MPPT techniques, the most popular ones are those known as Perturb and Observe (P&O) [2]–[4] and those based on the Incremental Conductance (InCond) [5], [6]. The aim of this work is to study a photovoltaic standalone pumping system, optimizing not only the efficiency of the photovoltaic conversion, using a maximum power point tracking algorithm, but also minimizing the losses in the induction motor that drive the centrifugal pump. Such thing has not yet gained much attention to this specific application and it consists in operating the motor at its maximum efficiency which, in association with the MPPT, renders the whole system even more efficient. The principle behind this technique is quite simple and it is based on the control of stator voltage. It will be shown in section III that, under partial loads, instead of keeping the same Volts-Hertz ratio, a reduction on flux level can minimize losses and keep the machine close to its rated efficiency. Computer simulations were made to investigate how to work close to the minimum losses point and propose a minimum losses point tracking (MLPT) algorithm. Then, it was tested on the experimental system to quantify its improvement and a gain up to 10% of machines’s rated power was estimated. In such a system where energy price is one of the key points, this result is quite impressive. Both improvement techniques work with small perturbations on the steady state, which can cause misworking. To avoid conflict between them, which can lead to misworking, some care need to be taken and it will be discussed in section IV. The development and implementation of a photovoltaic pumping system are presented in the next section. The motor losses reduction theory is exposed and studied in section III. Section V brings some experimental results, obtained with a supervisory system, which measures and stores the most important variables of the system, and real-time data from the Digital Signal Controller (DSC), which provide means for better analyzing and designing both algorithms. II. PV The pumping system (Fig. 2) is composed by 12 solar panels of 120 Wp each, a conventional 1 hp three-phase squirrel-cage induction motor (IM) and a centrifugal pump (CP). A standard three-leg IGBT PWM inverter converts DC voltage of PV panels into AC voltage to feed the motor. The inverter control allows varying the motor speed (open loop) and thus, modulating the load to the panels and tracking the maximum power point of operation of the array. The inverter is controlled by a low cost DSC manufactured by Freescale, model MC56F8013. The measured variables available to the DSC are Ipv , Vpv , Ia and Ib , which are the array’s current and voltage, and two of three motor’s line current, respectively. As this system main advantages are: • • • • • • Ipv PUMPING SYSTEM DESCRIPTION The converter topology results in a high efficiency converter, with losses less than 10%. Also, as it uses a standard inverter topology, there are few options of modules at an affordable cost; It uses induction motors and standard centrifugal pumps, which are cheaper, more rugged and with a better maintenance frequency/cost than other technological options; It is a modular system: it can be increased gradually by adding new panels; It does not require energy accumulators, which imply in lower initial investment and low maintenance cost, as well as reduced losses, complexity and weight; It works automatically with no need for operators; The system design concepts are general and not specific for utilized components, which allow great flexibility of use. A. MPPT method As mentioned in the previous section, several MPPT methods have been proposed in the literature, applying several different techniques as, for example, fuzzy logic and neural networks. In order to evaluate the quantity of papers published on this area, [1] lists 91 articles in its references, all of them proposing either new approaches or improvement on the existing ones. These approaches vary also with respect to their complexity, used sensors, convergence rate, steady-state error and cost, among others. The algorithm adopted in this work was the incremental conductance, which computes the inclination of curve Vpv x Ppv using measurements of array’s voltage and current (1), establishing whether the systems is at the current source region, voltage source region or at the maximum power point. PV Ipv Vpv Icc Ic C Inverter Fig. 2. a Ia b c Ib Ic IM Simplified diagram of the system. CP MPPT Vref + - C(s) ωe G(s) Icc - + Ic 1 Cs Vpv Vpv Fig. 3. Block diagram of the MPPT control. TABLE I L OGIC OF THE MULTI - CRITERION ALGORITHM [9] Measurements ∆Ppv ∆Vpv ∆Ipv <0 <0 <0 <0 <0 ≥0 <0 ≥0 <0 >0 <0 ≥0 >0 ≥0 <0 >0 >0 <0 =0 – – Vref ↑ ↑ ↓ ↓ ↑ ↓ ↔ ∂P ∂I = I + V. ∂V ∂V State of the PV Irradiance ↓ Current source Voltage source Voltage source Current source Irradiance ↑ MPP (1) This is one of first proposed MPPT algorithms and it has been chosen for some nice basic features as convergence rate, simplicity, low number of sensors, low cost and low steady-state error. These advantages, when compared to other strategies, are described in [1], [7], [8]. The main drawback of this approach when applied to pumping systems, as described in [9], is its lack of stability when fast irradiance changes occur. This was also verified in our experimental setup and solved by adding an inner voltage control loop. The general block diagram of MPPT control is depicted in Fig. 3, where G(s) represents the system dynamics and ωe is the stator voltage angular frequency. This is a more formal and natural approach than that adopted by [9]. It allows a better design of the controller and can avoid problems such as those noticed in [10]. When implementing the incremental conductance approach using a fixed point DSC, there is a limitation computing the division Ipv /Vpv which can, eventually, imply in unacceptable errors. This potential problem is avoided by using a multicriterion logic described in Table I. The perturbation frequency and its amplitude were both chosen by pratical tests. The first one was easier to chose, as it is possible to check the control loop response time by experimental means. The frequency should be as fast as the controller allows, or in other words, as soon as the panels voltage has stabilized at the next set point, another perturbation can be done. The amplitude is more hard to determine and a more formal approach must be chased, which is beyond of this article scope. As known, it may not be to high, as it will result a poor steady state response, neither to low, which causes problems with quick changes in irradiance. The values adopted to those variables were 2Hz and 1% of the open circuit voltage (Voc ). Is B. Automation A crucial concern in an autonomous pumping system, which can be installed several kilometers away of a nearby assistance, is its robustness and the automation of its operation. Nevertheless, even when the system is located in inhabited places, as in isolated communities, the availability of technicians with the necessary skills to operate and maintain such systems is quite unlikely. Therefore, some strategies have been adopted, which will be described next. Given that the water column has to be overcome, there is a minimum motor speed that produces a significant water flow. Thus, the hydraulical and electrical systems determine a minimum frequency for the motor voltage supply. On the other hand, if the irradiance decreases below a certain level and the available power is not sufficient to drive the motor, the voltage in the panels will drop and the system will be turned off due to under voltage protection. This procedure will wait for the measured voltage to be inferior to a pre-established value (50% of Voc ) for at least 10 cycles of the PWM and then, it will withdraw the pulses from the inverter. After a short interval, a new start should be commanded. The aim of starting procedure is to ensure that the system operates only when it is possible to pump water. For that sake, a voltage/frequency ramp is applied to the motor up to the minimum operating point, when the DC link voltage (Vpv ) is used as the first reference value and the MPPT algorithm assumes control. If during the motor acceleration the DC voltage collapses, this means that the irradiance is very low. So, a longer period of time has to be waited before a new start is tried. If, at any moment, Vpv stays for more than 10 PWM cycles below its inferior admissible limit, for any possible reason as, for instance, some perturbation that the control was not capable to reject, or the rotor blocking, etc., then the system stops and a new start procedure is immediately commanded. As the night comes, the panel voltages reduce considerably. Then, the controller detects that there is no irradiance anymore and starts the hibernation mode. This mode lasts for 10 hours, after which it starts to measure the DC link voltage again. In the case this voltage stays below 20% of Voc , new tests are made, every 15 minutes. If the DC link voltage is over 50% of Voc , the system enters in the active mode and starting procedures as described above are initiated. III. I MPROVEMENT OF MOTOR EFFICIENCY A. Theory The principle of flux weakening applied to induction motors under partial loads to improve induction machines efficiency dates back, at least, to 1983, when Rowan and Lipo have studied its effect on a silicon-controlled rectifier (SCR) voltage controller drive system [11]. From that time on, many authors have been studying this matter [12]–[16]. Nevertheless, it is not usual to be applied on a photovoltaic pumping system, where it can be really advantageous, as in this kind of system partial load is rule (as only at few moments the motor works Rs IsT σLs + + Vs Er - - Fig. 4. Ife Isλ Rfe L m’ Rr’.ωe sωe Induction motor’s vetorial equivalent circuit at rated load) and, moreover, a few watts on savings represents great amount of money. To understand why it is possible to improve motor’s efficiency, let us analyze the induction machine’s equivalent circuit of Fig. 4, which represents the core losses as a equivalent shunt resistance (Rf e ) [17]. It is known that the eletromagnetic torque (Te ) is (2) Te = 3(P/2).(Lm /Lr )λr IsT , and the rotor flux (λr ) is (3) λr = Lm .Isλ , where Lm is the mutual inductance, Lr is the rotor self inductance, P is the number of pole pairs, IsT and Isλ are the torque and flux currents, as it is shown in Fig. 4. For a given load condition, there is an infinite number of pairs λr –IsT which keep the system in its equilibrium. As Te is constant, a rotor flux reduction will lead to a increase in IsT and vice-versa. As λr is proportional to Isλ and the motor losses (PLoss ) can be expressed by (4), there must be a 2 flux which minimizes Ploss , i. e., a compromise between IsT 2 losses and Isλ ones. PLoss = 3Rs .|Is |2 + 3Rf e |If e |2 + 3.Rr0 |IsT |2 (4) Using (2) and (3), we can write the losses as a function of the motor’s parameters, angular frequency (ωe ), load torque (TL ) and rotor flux, (5). PLoss " Rs =3 + L2m # (Rs + Rf e ) . |λr |2 + Rf2 e 2 (Rs + Rr0 ) 2 Lr 1 + . .TL 3 P Lm |λr |2 2 R s Lr . ωe .TL . +2 . P R f e Lm ωe L m Lr 2 (5) Notice that (5) is a convex function where it’s minimum can be easily obtained as a function of the optimum flux: p |λro | = kλo (ωe ) TL (6) where kλo is equal to v u 2 u (Rs +R0r ) 2 Lr u . 3 P Lm 4 . kλo (ωe ) = u u 2 t (Rs +Rf e ) ω e Lm Rs 3 L2 + . R2 Lr m fe (7) B. Obtaining efficiency improvements As shown, loss reduction can be achieved by voltage control as function of machine’s load and frequency. The main question now is: how to determine the correct voltage? Two different strategies have been used in the literature: machine’s parameters are estimated and some relation, as in (6), is used to determine which voltage is to be applied; or an adaptative scheme is adopted. One very common strategy is the input power minimization. This approach works quite well for constant loads, but in those which torque decreases with speed, the minimum input power occurs when the motor is not moving at all. A mathematical model including core saturation and losses was used in order to evaluate the alternatives of input variables that could, in some way, be related with the maximum efficiency point. For this model, a test procedure of the motor was carried out, following IEEE-Std 114-2004. Windage and stray-load losses were not considered in this work, because of its hardness of measure and modeling [19]. Fig. 5 and Fig. 6 show the machine input variables and also the power output and efficiency for two electrical frequencies, 50Hz and 40 Hz, respectively. The load was modeled as in (8), where T60Hz is the pump nominal torque in per unity. The circles shows maximum or minimum points. As it was already expected, slip increases with voltage reduction, and the input and output powers are strictly decrescent, what makes the minimum input power approach impracticable. In both figures, the minimum current point is located very close to the maximum efficiency point. TL [pu] = T60Hz .ωe [pu]2 (8) This closeness gives us a clue that the current minimization strategy might be a good way of tracking the maximum efficiency point. The results presented in Fig. 5 and Fig. 6 were made with T60Hz equal to 1 pu. This might not be the rule, as the pump may not have been correctly chosen. To validate the minimum current strategy, the maximum efficiency and the efficiency at minimum current point were calculated for a wide range of the parameter T60Hz , in (8), from 0.2 pu to 1.2 pu, at the frequency range of interest, from 0.62 pu to 1.1 pu. The graph of Fig. 7 shows the MLPT’s relative efficiency for the mentioned range. In this graph, a perfect quadratic load would be represented as a straight vertical line. The result is far beyond expected, with a maximum “tracking error” less than 0.03%! In other words, the worst case is only 0,03% worse then the best one. Power [pu]; Current [pu]; Effiency; Power factor 1.6 1.4 input power output power efficiency input current power factor 1.2 1 0.8 0.6 0.4 0.2 0 0.3 0.4 0.5 0.6 Stator voltage [pu] 0.7 0.8 0.9 Fig. 5. Effect of voltage variation on efficiency, power output and input variables (ωe = 50Hz). Power [pu]; Current [pu]; Efficiency; Power factor Equation (6) shows us that there is an unique flux for which the machine’s efficiency is maximum. In fact, the efficiency is not only at a maximum, but is constant for all loads [18] in a given frequency. Although in this approach the core saturation was not taken into account, for the sake of simplicity, it is clear that, as the equivalent inductance is increased with the decrease of flux, the effect in the current reduction is greater than it would be without saturation and the loss reduction is even greater. 1.6 input power output power efficiency input current power factor 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0.2 0.3 0.4 0.5 Stator voltage [pu] 0.6 0.7 Fig. 6. Effect of voltage variation on efficiency, power output and input variables (ωe = 40Hz). The results presented are all for steady state, which is a reasonable assumption, given that both the solar irradiance and the hydraulic dynamics are slow if compared with the electromagnetic one. The tracking of the minimum current point is quite similar to MPPT methods. Here, again, a fast track response, low steady state error, low complexity and robustness are aimed. As a first approach, the perturb and observe algorithm was chosen. It isn’t the best method, its drawback are largely known, but, as it is very simple, it can be used to test the maximum efficiency strategy feasibility. The results obtained in the system described in section II are shown and discussed in section V. IV. MPPT AND MLPT DECOUPLING Although the efficiency improvement methods use small perturbations in the steady state, both can be used together without compromising each other performance. This is a matter of great concern, as unwanted interferences can cancel the methods benefits or even have a perverse effect, causing system’s performance to become worse. Thus, it surely needs further discussion. The aim here is to analyze what would be the crossed effects of the MLPT’s perturbation on the panel’s variables and of the MPPT ones on the motor input. 1.1 1 1.05 e Electric angular frequency, ω [pu] 0.9995 1 (a) 0.95 0.999 0.9 0.9985 0.85 (c) 0.8 (b) 0.998 0.75 0.9975 0.7 0.65 0.2 0.997 0.4 0.6 0.8 1 Normalized torque with variable quadratic base, T 60Hz Fig. 7. [pu] 1.2 Minimum current point relative efficiency Let us start with the first case. The MLPT changes the motor voltage searching for the minimum current point. If the stator voltage is reduced and the frequency is kept constant, motor’s slip would increase, pump speed decreases and the output power also would be reduced. As a consequence, the load seen by the panels would change and their operation point is also perturbed, but not by the MPPT algorithm. This would be true if the DC link voltage was not controlled. But, as it is controlled, frequency increases to maintain DC voltage constant, the operation point of the panels does not change, and the final result is an increase in pump speed and output power, as expected. The second case is harder to analyze. Throughout the whole motor losses minimization analysis, frequency was treated as constant, although it isn’t. It is, in fact, the main control variable of the entire system and it is affected by three factors: solar irradiance variation, set point changes of DC link voltage (caused by the MPPT) or the MLPT itself. From those three types of variations, irradiance changes have the greatest potential of misleading the MLPT method, for at least two reasons: there is no control on solar irradiance and it can change abruptly. Two alternatives were taken into account on how MLPT actuates to modify stator voltage. It can be either acting directly upon this variable (Fig. 8) or in an indirect way, acting on the volts-hertz ratio (Fig. 9). When irradiance increases, the stator current increases as well and the sign of perturbation changes every MLPT cycle (Fig. 8(a) and Fig. 9(a)). When it decreases, current follows it and the sign of perturbation is kept constant throughout the entire fall process (Fig. 8(b,c) and Fig. 9(b,c)). The indirect method does not cause a side cross (from dIs /dVs > 0 to dIs /dVs < 0 or the reverse) and almost maintains the relative operation point on cases (a) and (b). In fact, a deeper analysis shows that when the indirect method is used, the variable that is controlled (and kept constant between steps) is related to the flux. On the other hand, the direct method controls voltage only, and that is why the first is better than the second. Stator voltage [pu] Fig. 8. Behaviour of MLPT algorithm under irradiance change, direct MLPT (c) (b) (a) Stator voltage [pu] Fig. 9. MLPT Behaviour of MLPT algorithm under irradiance change, indirect V. E XPERIMENTAL RESULTS Experimental results were obtained, which confirm the improvement of the stand-alone photovoltaic pumping system in terms of its efficiency when the extra algorithm is added to the MPPT technique. Preliminary results are shown, which give a quantitative idea of this improvement. The evaluation of the MLPT algorithm was made with the inverter supplied by the grid, in order to keep the power constant and render the analysis of the approach easier. The procedure was to, keeping a constant frequency (f ), run the system without and with the MLPT on. Fig. 10 shows the result obtained with f = 45Hz. Notice that, when the algorithm is turned on, both the input power and the speed (and output power) decrease. It was quite expected, since the slip frequency increase is not compensated by a synchronous frequency increase. It is important to have an idea on how the output power changes with speed. In the used pump, small perturbations on an operation point leads to almost no change on torque, so Po = k.ωR , where Po is the power output, ωR the rotor speed and k some proportionality constant. It can be easily shown that for this load dPo dωR = , Po ωR (9) what means that if the rate of change in the power input 0.12 1.4 P = k.ω Pin [pu] 1.2 o Vs [pu] 1 100 200 300 400 500 600 700 800 900 Speed [rpm] filtered [rpm] 1320 Power saved [pu] 0 1340 0.08 0.06 0.04 0.02 0 0.65 1300 Fig. 10. R s 0.6 1280 o I [pu] 0.8 0.4 R P = k.ω2 0.1 0 100 200 300 400 500 Time [s] 600 700 800 900 Fig. 11. 0.7 0.75 0.8 0.85 Frequency [pu] 0.9 0.95 1 Estimative of the power saved by the MLPT Test of minimum losses point tracking algorithm for f = 45Hz (dPi /Pi ) is greater than the speed change (dωR /ωR ), some efficiency improvement is achieved. In order to be more conservative, the pump should be modeled as quadratic function 2 of the speed (Po = k.ωR ), and efficiency would improve if dPi /Pi > 2.dωR /ωR . Notice that the slip frequency variation is very small and limited. In fact, when reducing the motor losses through the MLPT, while keeping the PV array at its point of maximum power, the saving of losses of the motor is converted in useful power at the pump. This is done as follows: the DC voltage control increases the stator frequency with the motor losses reduction due to the MLPT. A higher frequency applied to the motor causes an increase in the rotor speed an in the power delivered to the load. Fig. 11 shows an estimative of the power saved for the two models of the load. It is expected that the savings are somewhere between the two lines. VI. C ONCLUSION The presented work aims the optimization of a stand-alone solar pumping system. The improvements proposed here can ease the path of this kind of system to popularization, reducing the initial costs and its payback time. The proposed motor losses minimization algorithm is quite similar to a known MPPT, as neither of them require sensors, search for maximum efficiency operation points and use small perturbations to achieve their goals. It was shown that with the proposed algorithm it is possible to reduce the motor losses and have power savings up to 10% of input power, maximizing the converted solar energy. The experimental results show the feasibility of this strategy and confirm the theory. R EFERENCES [1] T. Esram and P. L. Chapman, “Comparison of photovoltaic array maximum powrer point tracking techniques,” IEEE Transactions on Energy Conversion, vol. 22, no. 2, pp. 439–449, Jun 2007. [2] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “Optmizing of perturb and observe maximum power point tracking,” IEEE Trans. on Power Electronics, vol. 20, no. 4, pp. 963–973, Jul 2005. [3] C. Hua and J. R. Lin, “Dsp-based controller application in battery storage photovoltaic system,” In Proc. IEEE IECON ’96, pp. 1705–1710, 1996. [4] N. S. Souza, L. A. C. Lopes, and X. Liu, “An intelligent maximum power point tracker using peak current control,” In Proc. IEEE PESC’05, pp. 172–177, 2005. [5] C. Pan, J. Chen, C. Chu, and Y. Huang, “A fast maximum power point tracker for photovoltaic power systems,” In Proc. IECON’99, pp. 390– 393, 1999. [6] T. Kim, H. Ahn, S. K. Park, and Y. Lee, “A novel maximum power point tracking control for photovoltaic power system under rapdily changing solar radiation,” IEEE int. Symp Ind Electronics, pp. 1011–1014, 2001. [7] K. C. Oliveira, M. C. Cavalcanti, G. M. S. Azevedo, and F. A. S. Neves, “Comparative study of maximum power point tracking techniques for photovoltaic systems,” In Proc. VII INDUSCON, Aug 2006. [8] D. P. Hohm and M. E. Ropp, “Comparative study of maximum power point tracking algorithms using an experimental, programmable, maximum power point tracking bed,” IEEE, pp. 1699–1702, 2000. [9] G. Heng, X. Zheng, L. You-Chun, and W. Hui, “A novel maximum power point traking strategy for stand-alone solar pumping system,” IEEE/PES Transmission and Distribution Conference and Exhibition, pp. 1–5, 2005. [10] S. Zhang, Z. Xu, Y. Li, and Y. Ni, “Optimization of mppt step size in stand-alone solar pumping system,” Power Engineering Society General Meeting, p. 6 pp, Jun 2006. [11] T. A. Lipo, “A quantitative analisys of induction motor performance improvement by scr voltage control,” IEEE Trans. on Industry Applications, vol. IA-19, no. 4, pp. 545–553, July/August 1983. [12] D. S. Kirschen, D. W. Novotny, and W. Suwanwisoot, “Minimizing induction motor losses by excitation control in variable frequency drives,” IEEE Trans. on Industry Applications, vol. 20, pp. 1244–1251, sep/oct 1984. [13] K. Matsuse, S. Taniguchi, T. Yoshizumi, and K. Namiki, “A speedsensorless vector control of induction motor operating at high efficiency taking core loss into acount,” IEEE Trans. on Industry Application, vol. 37, no. 2, pp. 548–558, Mar/Abr 2001. [14] M. N. Uddin and S. W. Nam, “Adaptive backstepping based online loss minimization control of an im drive,” Canadian Journal of Electrical and Computer Engineering, vol. 32, pp. 97 – 102, 2007. [15] M. Tsuji, S. Chen, T. Kai, E. Yamada, S. Hamasaki, and A. D. Pizzo, “A precise torque and high efficiency control for q-axis flux-based induction motor sensorless vector control system,” SPEEDAM 2006, pp. 990 – 995, May 2006. [16] H. Sepahvand and S. Farhangi, “Enhacing performance of a fuzzy efficiency optimizer for induction motor drives,” PESC ’06, pp. 1 – 5, June 2006. [17] E. Levi, A. Lamine, and A. Cavagnino, “Impact of stray load losses on vector control accuracy in current-fed induction motor drives,” IEEE Trans. on Energy Conversion, vol. 21, pp. 442–450, Jun 2006. [18] T. W. Jian, N. L. Schmitz, and D. W. Novotny, “Characteristic induction motor slip values for variable voltage part load performance optimization,” IEEE Trans. on Power Apparatus and Systems, vol. PAS-102, no. 01, pp. 38–46, Jan/Feb 1983. [19] K. J. Bradley, W. Cao, and J. Arellano-Padilla, “Evaluation of stray load loss in induction motors with a comparison of input-output and calorimetric methods,” IEEE Trans. on Energy Conversion, vol. 21, no. 3, pp. 682–689, Sep 2006.