- Open Access
- Total Downloads : 910
- Authors : G.Srinivas , Dr.S.Tarakalyani
- Paper ID : IJERTV2IS90172
- Volume & Issue : Volume 02, Issue 09 (September 2013)
- Published (First Online): 07-09-2013
- ISSN (Online) : 2278-0181
- Publisher Name : IJERT
- License: This work is licensed under a Creative Commons Attribution 4.0 International License
Optimal Control to Sensor less Vector Control of Induction Motor using AI Techniques
Optimal Control to Sensor less Vector Control of Induction Motor using AI Techniques
-
rinivas Dr.S.Tarakalyani
Research Scholar, JNTUCEH Professor, JNTUCEH
Hyderabad, A.P, India Hyderabad, A.P, India
Abstract- this paper deals with a new design optimal control to Sensor less Vector Control of Induction Motor using Artificial Intelligent (AI) Techniques. This study discusses a method to control the speed of three phase induction motor based on sensor less Control using Neural Network (NN). With this method, it is expected to produce accurate steady-state speed response for the motor parameters (electromagnetic torque and stator flux) can be set directly. The analysis is done with PI, Fuzzy & ANN controllers. An optimal SIMULINK/MATLAB model has been designed in order to achieve the speed control of three phase induction motor based on sensor less control method.
Index Terms Artificial Intelligent (AI), ANN, speed estimator, induction motor, and efficiency optimization,
-
INTRODUCTION
Induction motor is the work horse in industry due to its rigid construction & can work under all conditions of environment. But due to the factor that flux & torque cannot be controlled individually as the stator current is a combination of both it is not popular like D.C. Motor. But with the development of power electronics & Vector control concept the three phase stator current can be resolved into two phase components by orthogonal transformation by using Clarkes transformation& to rotor reference frame by parks transformation. To do this the position of flux vector is important. This position of flux vector can be found by direct & indirect methods where direct method employs sensors incorporated in stator which adds to cost, size & induction of harmonics. Hence in indirect control this flux vector can be found by machine parameters & modeling equations governing its performance. Hence sensor less vector control has gained importance. Basically there are various methods of indirect vector control of which Kalman filter, MRAS, sliding mode observer are in major use in earlier days, and hence these methods are prone to numerical & steady errors due to large calculations involved. Hence with the development of softwares like Matlab/Simulink, & computer methods like fuzzy logic,
The present paper deals with calculation of torque, speed of induction motor without optimization controller & comparing it with after installing controller using ANN approach.
-
MODELING OF INDUCTION MOTOR
-
Dynamic Modeling of induction motors
This section presents the dynamic model of the induction motor as shown in the below fig.1, it is derived by transforming the three-phase quantities into two phases direct and quadrature axes quantities. The equivalence between the three-phase and two-phase machine models is derived from the concept of power invariance [16].
Fig.1.Block diagram of the dynamic model of the induction motor
Electromagnetic Torque:
neural networks the complications have been resolved.
= 3
(1)
The Indian power sector has come long way in power generation from 1300MW capacity during independence to
2 2
102907MW at present. However in spite of governments plans, the present power availability is not good enough to cater to the needs of the country, as there is a peak shortage of the power of around 10,000MW (13%) and 40,000 million units deficit (7.5%). Unless the system efficiency improves in terms of technical improvements, the crisis will still continue. Energy savings possible due to some major energy equipments such as transformers, motors etc.
The dynamic equations of the induction motor in synchronous reference frames can be represented by using flux linkages as variables. This involves the reduction of number of variables in dynamic equations, which greatly facilitates their solution. The flux-linkages representation is used in motor drives to highlight the process of the decoupling of the flux and torque channels in the induction
machine. The stator and rotor flux linkages in synchronous reference frame are shown in equations (2)-(7)
=1 – 2
(16)
qs
= Lsiqs
+ Lm iqr
(2)
ds
= Lsids + Lm idr
(3)
qr
= Lriqr + Lm iqs
(4)
dr
= Lridr + Lm ids
(5)
= +
(6)
= +
(7)
ds
= Lsids + Lm idr
(3)
qr
= Lriqr + Lm iqs
(4)
dr
= Lridr + Lm ids
(5)
= +
(6)
= +
(7)
-
-
SENSOR LESS VECTOR CONTROL TECHNIQUES
-
Sensor less vector control techniques by using fuzzy
control
-
State Space model of induction motor
The space phasor model of the induction motors can be presented in state space equations from previous equation, so it can be expressed in the synchronously rotating d-q reference frame as shown in equations (8) to (16).
= + (8)
= [ ] (9)
= [ ] (10)
Fig. 2. The indirect vector controlled induction motor with efficiency optimization controller block diagram
The principle of efficiency optimization control with rotor flux programming at a steady-state torque and speed condition is explained in Fig.: 2 The rotor flux is decreased by reducing the magnetizing current, which ultimately results in a corresponding increase in the torque current (normally by action of the speed controller); such that the developed torque remains constant. As the flux is decreased, the iron loss decreases with the attendant increase of copper loss.
=
3 0
+
(11)
Fig.3.Efficiency optimization control block diagram
The above figure3 explains the fuzzy efficiency controller
B = 0 3 (12)
operation. The input dc power is sampled and compared
0 0 with the previous value to determine the increment Pd. In
0 0 addition, the last excitation current decrement (L ids) is reviewed. On these bases, the decrement step of ids* is generated from fuzzy rules through fuzzy inference and
1 = (1)
(13)
defuzzification. The adjustable gains Pb and Ib, generated by
2
=
2
1
(14)
scaling factors computation block, convert the input variable and control variable, respectively, to per unit values so that a single fuzzy rule base can be used for any torque and speed condition. The input gain Pb as a function of machine speed wr can be given as Pb=a wr+b Where the coefficients a and
b were derived from simulation studies. The output gain Ib is
3 = (1)
computed from the machine speed and an approximate estimate of machine torque Te Ib=c1 wr-c2 Te+c3 Again, the
appropriate coefficients c1, c2, and c3 were derived from simulation studies. In the absence of input and output gains, the efficiency optimization controller would react equally to a specific value of Pb resulting from a past action ids*(k- 1), irrespective of operating speed. Since the optimal efficiency is speed dependant, the control action could easily be too conservative, resulting in slow convergence, or excessive, yielding an overshoot in the search process with possible adverse impact on system stability. As both input and output gains are function of speed, this problem does not arise. The above equation also incorporates the a priori knowledge that the optimum value of ids* is a function of torque as well as machine speed. In this way, for different speed and torque conditions, the same ids*(p.u) will result in different ids*, ensuring a fast convergence. One additional advantage of per unit basis operation is that the same fuzzy controller can be applied to any arbitrary machine, by simply changing the coefficients of input and output gains.
The membership functions for the fuzzy efficiency
controller are shown below. Due to the use of input and output gains, the universe of discourse for all variables are
()/()
N
P
PB
PM
NM
PM
PS
NS
PS
PS
NS
ZE
ZE
ZE
NS
NS
PS
NM
NM
PM
NB
NB
PB
Rules base for Fuzzy Efficiency Controller:
As the excitation current is decremented with adaptive step size by the fuzzy controller, the rotor flux dr will decrease exponentially which is given by equation (17)
normalized in the [-1, 1] interval. It was verified that, while
=
(17)
the control variable ids*, required seven fuzzy sets to
provide good control sensitivity, the past control action L
ids* (i.e. ids* (k – 1)) needed only two fuzzy sets, since the main information conveyed by them is the sign. The small overlap of the positive (P) and negative (N) membership functions is required to ensure proper operation of the height defuzzification method, i.e., to prevent indeterminate result in case L ids* approaches zero. The rule base for fuzzy control is given below. The basic idea is that if the last control action indicated a decrease of dc link power, proceed searching in the same direction and the control magnitude should be somewhat proportional to the measured dc link power change. In case the last control action resulted in an increase of Pd (Pd > 0), the search direction is reversed, and the ids*, step size is reduced to attenuate oscillations in the search process.
Fig.4a,b&c Membership functions for efficiency controller change of DC link power (Pd(pu)) ,Last change in excitation current (Lids*(pu)) & Excitation current control increment(ids*(pu)).
Depending on the member ship functions for efficiency
Where r = Lr/Rr is the rotor time constant and Lm the magnetizing inductance. The decrease of flux causes loss of torque, which normally is compensated slowly by the speed control loop. Such pulsating torque at low frequency is very undesirable because it causes speed ripple and may create mechanical resonance. To prevent these problems, a feed forward pulsating torque compensator has been proposed.
Under correct field orientation control, the developed torque is given by equation (18)
= (18)
For an invariant torque, the torque current Iqs, should be controlled to vary inversely with the rotor flux. This can be accomplished by adding a compensating signal
Iqs* to the original Iqs* to counteract the decrease in flux
dr (t) where t [0, T] and T is the sampling period for efficiency optimization control. Let iqs (0) and dr (0) be the initial values for iqs and dr, respectively, for the k-th step change of ids*. For a perfect compensation, the developed torque must remain constant, and the following equality given by equation (19) holds good.
[dr 0 + dr t ] iqs 0 + iqs t = dr 0 iqs 0(19)
Solving for Iqs (t) yields
controller change of dc link power,Last change in excitation current,& excitation current control increment shown in Fig.4a,b&c the rule base for fuzzy controller is framed and
= 0
0 + ()
(20)
is shown in table 1
Where dr (t) is governed by above equation with, substituted for ids*. To implement such compensation, above equations are adapted to produce Iqs (t), using flux estimate dr and command in Iqs* place of actual signals. A good approximate solution for Iqs(t) can be obtained by
replacing the denominator of the above equation by its steady-state value estimate dr (t).In this case the compensation can be implemented in two steps as shown in Fig. 4. First, the value for the compensated torque current step is computed by discrete given by equation (20) as
summation function is found by multiplying each component of the i vector by the corresponding component of the w vector and then adding up all the products. Input 1 = i1 * w1, input 2 = i2 * w2 etc., are added as input 1 + input 2 + + input n.
The result of the summation function, the weighted sum,
=
1 1 (21)
is transformed to a working output through an algorithmic process known as the transfer function. In the transfer function the summation total can be compared with some threshold to determine the neural output. If the sum is
Next, the current step is processed through a first order low pass filter of rotor time constant, and then added to the previous compensating steps. This current is added to the original speed loop generated current Iqs* so that, at any instant, the product Iqs dr remains essentially constant
-
Sensor less vector control techniques by using Artificial Neural Network method
An Artificial Neural Network (ANN), often just called a "neural network" (NN), is a mathematical model or computational model based on biological neural networks. It consists of an interconnected group of artificial neurons and processes information using a connectionist approach to computation. In most cases an ANN is an adaptive system that changes its structure based on external or internal information that flows through the network during the learning phase.
Fig. 5. Basic Neuron network model
An artificial neuron consists of five major components, They are Weighing factors, summation function, transfer function, scaling& limiting function, output function. A neuron usually receives many inputs simultaneously. Each input has its own relative weight which gives the input the impact that it needs on the processing elements summation Weights are a measure of an inputs connection strength. These strengths can be modified in response to various training sets and according to a networks specific topology or thought its learning rules.
The first step in a processing elements operation is to compute the weighted sum of all of the inputs. Mathematically, the inputs and the corresponding weights are vectors which can be represented as (i1, i2 in) and (w1, w2 wn). The total input signal is the dot, or inner, product of these two vectors. This simplistic
greater than the threshold value, the processing element generates a signal. If the sum of the input and weight prducts is less than the threshold, no signal (or some inhibitory signal) is generated. Both types of response are significant. The threshold, or transfer function, is generally non-linear. Linear (straight-line) functions are limited because the output is simply proportional to the input. Linear functions are not very useful.
After the processing elements transfer function, the result can pass through additional processes which scale and limit. This scaling simply multiplies a scale factor times the transfer value, and then adds an offset. Limiting is the mechanism which insures that the scaled result does not exceed an upper or lower bound.
Each processing element is allowed one output signal which it may output to hundreds of other neurons. This is just like the biological neuron, where there are many inputs and only one output action. Normally, the output is directly equivalent to the transfer functions result. Some network topologies, however, modify the transfer result to incorporate competition among neighboring processing elements. Neurons are allowed to compete with each other, inhibiting processing elements unless they have great strength. Competition can occur at one or both of two levels. First, competition determines which artificial neuron will be active, or provides an output. Second, competitive inputs help determine which processing element will participate in the learning or adaptation process.
a) ANN Controller for induction motor
Fig 6 ANN Controller for induction motor
Fig. 6 shows the control mechanism. If either the estimated flux or torque deviates from there reference more than allowed tolerance, the transistors of the variable frequency drive are turned off and on in such away that the flux and torque will return in their tolerance band as fast as possible.
Fig. 7. Simulink block diagram of ANN Controller switching table Figure 7 shows the switching table for ANN controller at which the controller varies various parameters to get steady state response.
Fig.8 . The block diagram of induction motor with ANN Controller for sensor less vector control.
The above figure8 shows the block diagram of induction motor incorporating ANN controller for sensor less vector control. The drive uses a PWM IGBT inverter fed from a DC source. The speed error is processed through a neural network. Neural Network is used to adjust the motor parameters when the induction motor speed reference changes and also to produce accurate steady-state speed response.
Fig. 9. Simulink block diagram of the multi-layer neural
network
Fig. 9 shows multilayer neural network. A network can have several layers. Each layer has a weight matrix W, a bias vector b, and an output vector the layers of a multilayer network play different roles. A layer that produces the network output is called an output layer. All other layers are called hidden layers. Multiple-layer networks are quite powerful. For instance, a network of
two layers, where the first layer is sigmoid and the second layer is linear, can be trained to approximate any function
Fig.10.Weights sub block
Fig.10 .shows A neuron usually receives many inputs simultaneously. Each input has its own relative weight which gives the input the impact that it needs on the processing elements summation Weights are a measure of an inputs connection strength
Fig.11. inputs for a neuron
Figure 11 shows the weights sub block & inputs for a neuron which are inter connected. A neuron usually receives many inputs simultaneously. Each input has its own relative weight which gives the input the impact that it needs on the processing elements summation Weights are a measure of an inputs connection strength.
-
-
SIMULATION RESULTS
A step increase in reference speed Wref produces a positive speed error. The drive accelerates at a permissible inverter current, producing maximum available torque, until the speed error is reduced to small value. The drive finally settles at a speed for which the motor torque balances the load torque. A step decrease in reference speed Wref produces a negative speed error. The speed command is set at the maximum negative value. The drive has fast response because the speed error is corrected at the maximum available torque. The error between the set speed and actual speed is calculated. In order to operate induction motor with variable speed, power converter devices are used with the power supply. A conventional way to control the induction motor speed is by using a PID controller, but tuning the PID controller to work under various speed references is very time consuming. The wave forms for speed ,Torque & voltages with PI Controller & ANN Controller are analyzed & it is true that ANN controller gives better response.
TABLE II
HP=5
Power rating of motor
V=440v
Voltage applied
F=50HZ
Frequency
N=1500RPM
Speed in RPM
P=4
No of poles
Rs = 0.406
Stator Resistance
Rr = 0.478
Rotor Resistance
Lls = 2.13mH
Stator Leakage Resistance
Llr = 2.13mH
Rotor Leakage Resistance
Lm = 49.4mH
Mutual Inductance
The simulation of the induction motor for sensor less vector control method was done with PI controller for a
simulation time of 2 seconds & the graphs for voltage ,speed & torque are as shown in figure12 to figure14..
Fig.12. The voltage wave form using PI controller
Fig.13. The speed wave form using PI controller
Fig.14 The Torque wave form using PI controller
The simulation of the induction motor for sensor less vector control method was done with Fuzzy controller for a simulation time of 2 seconds & the graphs for voltage ,speed & torque are as shown in figure15 to figure17.
Fig.15. The voltage wave form using fuzzy controller
Fig.16. The speed wave form using fuzzy controller
Fig.17 The torque wave form using fuzzy controller
The simulation of the induction motor for sensor less vector control method was done with ANN controller for a simulation time of 2 seconds & the graphs for voltage ,speed & torque are as shown in figure18 to 20
Fig.18. The voltage wave form using ANN controller
Fig.19. The speede wave form using ANN controller
Fig.20 The Torque wave form using ANN controlle
From the simulation studies of voltage wave forms it can be concluded that the voltage is smooth & ripple free with more pulses in a given interval. Hence by using ANN controller the voltage is maintained at highest value with less distortion hence better performance
From the analysis of rotor speeds it is observed that the maximum speed that can be attained is more with controller & it is found that this speed is maintained constant. Whereas without controller the maximum speed attained is less & after a short while speed drops drastically . Hence with ANN controller maximum speed can be increased & maintained constant.
The above simulation wave forms for torque show that the behavior of induction motor ie as torque decreases speed increases. The torque with controller is less distorted, almost all constant without jerky operation. Hence torque is improved & almost all constant with ANN controller.
-
CONCLUSION
-
In this paper a new MATLAB/SIMULINK model is proposed for speed control of three phase induction motor based on neural network approach. The proposed model gives accurate steady state response for the motor parameters (electromagnetic torque and stator flux). From the simulation results, it can be observed that the proposed modelgives accurate steady state response. But owing to the stability and learning capability of neural networks, the proposed method can be considered as better technique.
-
Takahashi, T. Noguchi, A new quick response and high-efficiency control strategy of induction motor, IEEE Trans. On IA, Vol.22, N°.5, Sept/Oct 1986, PP.820- 827.
-
M. Depenbrock, Direct self control (DSC) of inverter fed induction machine, IEEE Trans. Power Electronics, Vol.3, N°.4, Oct 1988, PP.420-829.
-
D. Casadei, F. Profumo, G.Serra and A.Tani,
FOC and DTC: Tox Viable Schemes for induction Motors Torque Control, IEEE Trans. Power Electronics. On PE, Vol.17, N°.5, Sept2002,
-
D. Casadei and G.Serra, Implementation of direct Torque control Algorithme for Induction Motors Based On Discrete Space Vector Modulation, IEEE Trans. Power Electronics. Vol.15, N°.4, JULY2002,
-
R.Toufouti ,H.Benalla, and S.Meziane Three- Level Inverter With Direct Torque Control For Induction Motor, World Conference on Energy for Sustainable Development: Technology Advances and Environmental Issues, Pyramisa Hotel Cairo – Egypt, 6 – 9 December 2004.
-
L. Fausett, "Fundamentals of Neural Networks: Architectures, Algorithm, and Applications", Prentice Hall, 2003.
-
Neural Networks, Fuzzy Logicand GeneticAlgorithms by S.Rajasekaran, G.A.Vijayalakshmi
-
Efficiency Optimization Control of A.C Induction Motors: Initial Laboratory Results , M.W.Turner, V.E.McCornick and J.G.Cleland, EPA Project EPA/600/SR- 96/008
-
Fuzzy logic based on-line efficiency Optimization of an indirect vectorcontrolled Induction motor drives, G.C.D. Sousa, B.K.Bose, J.G.Cleland, IEEE Transaction on Industrial Electronics, vol. 42, no.2, pp.192-198, April 1995.
-
Fuzzy logic drives.efficiency optimization of induction based improvements in induction motor drives. Juan M. Moreno-Eguilaz, M. Cipolla-Ficarra, P.J.da Costa Branco, Juan Peracaula,
-
A Fuzzy-Logic-Based Energy Optimizer for AC Motors John G. Cleland ,Vance E. McCormick and M. Wayne TurnerCenter for Digital Systems Engineering, Research Triangle Institute, Research Triangle Park, NC 27709 Pp: 1777-1784
-
An Efficient Controller for an Adjustable Speed Induction Motor Drive G. 0. Garcia, J. C. Mendes Luis, R.
M. Stephan, and E. H. Watanabe IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 5,
OCTOBER 1994 533 Pp: 533-539
-
Loss Minimization Control of an Induction Motor Drive Parviz amouri and Jimmie J. Cathey, Senior Member, IEEE IEEE TRANSACTIONS ON INDUSTRY
APPLICATIONS, VOL. 27. NO. 1, JANUARY
IFEBRUAR 1991 Pp: 32-37
-
. An Efficiency-Optimization Controller for Induction Motor Drives M.E.H. Benbouzid, N.S. Nait Said Pp: 63-64
-
Fuzzy Efficient-Optimization Controller for Induction Motor Drives F. Zidani, M.E.H. Benbouzid, D. Diallo Author Affiliation: University of Picardie Jules Verne, 7, Rue du Moulin Neuf, 80000 Amiens, France. pp:43-44
-
Loss Minimization of a Fuzzy Controlled Induction Motor Drive F. Zidani*, M.E.H. Benbouzid, Senior Member, IEEE and D. Diallo, Member, IEEE University of Batna, Algeria Centre de Robotique, dElectrotechnique et d Automatique University of Picardie Jules Verne7, Rue du Moulin Neuf – 80000 Amiens, France Pp: 629-633
Hori, Senior .