International Journal of Research and Reviews in Computer Science (IJRRCS) Vol. 2, No. 4, August 2011, ISSN: 2079-2557 © Science Academy Publisher, United Kingdom www.sciacademypublisher.com
959
DTC for Variable Speed Wind Turbine Driven Induction Generator Using ANN D.V.N. Ananth1, D. Himabindhu1, K.V. Ramana1, and V. Jagadeesh2 VITAM Engineering College, Vishakapatnam, India M.V.G.R. College of Engineering, Vizianagaram, India
1 2
Email:
[email protected],
[email protected],
[email protected],
[email protected]
Abstract – In this paper, Artificial Neural Network based wind turbine with induction generator system is analyzed to develop constant voltage, ripple free stator flux, faster response of the machine with load torque and generator speed. In General wind speed varies with time, season, location etc. Therefore, there is a need to study about variable-speed wind power generation system based on direct torque control (DTC) of an Induction generator (IG). The proposed system can achieve two main features, one is with change in the wind speed, output torque of IG changes, so output power changes. These changes can be achieved in short period, compared to previous ANN techniques. Secondly, if there is any requirement to change in the IG speed, the desired speed can be achieved at a faster rate. Hence this paper aims to present a predictive scheme to correct this time delay. Finally, simulation results show that the proposed DTC method effectively reduces the torque and flux ripples at low switching frequency, even under variable speed operation conditions. The result of DTC system simulation in MATLAB / SIMULINK shows that ANN speed recognition optimization algorithm has better tracking capability and fitness, as well as favorable static and dynamic properties. Keywords – Direct torque control (DTC), Wind turbine, Induction generator, Artificial neural network (ANN)
1.
Introduction
Direct Torque Control which can be applicable to asynchronous machines, permanent magnet machines etc, describes the way in which the control of torque and speed are directly based on the electromagnetic state of the motor. Similar to a DC motor, but contrary to the way in which traditional PWM drives use input frequency and voltage these machines can be controlled. DTC is the first technology to control the “real” motor variables of torque and flux. Because torque and flux are motor parameters that are being directly controlled, there is no need for a modulator, as used in PWM drives, to control the frequency and voltage. This, in effect, cuts out the middle man and dramatically speeds up the response of the drive to the changes in required torque. DTC also provides precise torque control without the need for a feedback device. To meet the major demands by the Industry, such techniques are necessary for • Better product quality which can be partly achieved with improved speed accuracy and faster torque control. • Less down time which means a drive that will not trip unnecessarily; a drive that is not complicated by expensive feedback devices; and a drive which is not greatly affected by interferences like harmonics and RFI. • Fewer products. One drive capable of meeting all application needs whether AC, DC or servo. That is a truly “universal” drive. • A comfortable working environment with a drive that produces much lower audible noise The benefit of DTC technology includes exceptional dynamic performance features, many of which are obtained without the need for an encoder or tachometer to monitor shaft position or speed:
Torque response: How quickly the drive output can reach the specified value when a nominal 100% torque reference step is applied. For DTC, a typical torque response is 1 to 2ms below 40Hz compared to between 10-20ms for both flux vector and DC drives fitted with an encoder. With open loop PWM drives the response time is typically well over 100ms. In fact, with its torque response, DTC has achieved the natural limit. With the voltage and current available, response time cannot be any shorter. Even in the newer “sensorless” drives the torque response is hundreds of milliseconds. With DTC, accurate torque control at low frequencies, as well as at full load torque at zero speed without the need for a feedback device such as an encoder or tachometer. With DTC, speed can be controlled to frequencies below 0.5Hz and still provide 100% torque right the way through to zero speed. Torque repeatability: DTC, without an encoder, can provide 1 to 2% torque [1] repeatability of the nominal torque across the speed range. This is half that of other open-loop AC drives and equal to that of closed-loop AC and DC drives. Motor static speed accuracy: Error between speed reference and actual value at constant load. For DTC, speed accuracy is 10% of the motor slip, which with an 11kW motor equals 0.3% static speed accuracy. With an 110kW motor, speed accuracy is 0.1% without encoder (open-loop). This satisfies the accuracy requirement for 95% of industrial drives applications. However, for the same accuracy from DC drives an encoder is needed. In contrast, with frequency controlled PWM drives the static speed accuracy is typically between 1 to 3%. So the potential for customer process improvements is significantly higher with standard drives using DTC technology. A DTC drive using an encoder with 1024 pulses/revolution can achieve a speed accuracy of 0.01% [1]. Dynamic speed accuracy: Time integral of speed deviation when a nominal (100%) torque speed is applied. DTC open-loop
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
dynamic speed accuracy is between 0.3 to 0.4% seconds [1]. This depends on the gain adjustment of the controller, which can be tuned to the process requirements. With other open-loop AC drives, the dynamic accuracy is eight times less and in practical terms around 3%sec. If we furnish the DTC controller with an encoder, the dynamic speed accuracy will be 0.1%sec, which matches servo drive performance. The stability down to zero speed is good and both torque and speed accuracy can be maintained at very low speeds and light loads. We have defined the accuracies as follows: Torque accuracy: Within a speed range of 2-100% and a load range of 10-100%, the torque accuracy is 2%. Speed accuracy: Within a speed range of 2-100% and a load range of 10-100%, the speed accuracy is 10% of the motor slip. Motor slip of a 37kW motor is about 2% which means a speed accuracy of 0.2%. Disadvantages of DTC: If several motors are connected in parallel in a DTC-controlled inverter, the arrangement operates as one large motor. It has no information about the status of any single motor. If the number of motors varies or the motor power remains below 1/8 of the rated power, it would be best to select the scalar control macro. The ANNs are capable of learning the desired mapping between the inputs and outputs signals of the system without knowing the exact mathematical model of the system. Since the ANNs do not use the mathematical model of the system, the same. The ANNs are excellent estimators in non linear systems [6] - [8]. Various ANN based control strategies have been developed for direct torque control of wind driven induction generator set to overcome the scheme drawback. In this paper, neural network flux position estimation, sector selection and switching vector selection scheme proposed, and ANN based speed controller used to reduce the current ripple by regulating the switching frequency, are proposed. The organization of this paper goes on in the following order. In Section II, it will be presented the basic concept of DTC for induction motor drive. In Section III and IV it will be described flux estimation algorithm and artificial neural networks and implementation of ANN to the DTC scheme. The simulation and experimental results will be presented in Section V and VI for the proposed scheme validation. In Sections VII, it will be presented the conclusions of this work.
960
There is no need to feed back any shaft speed or position
Figure 2.1. DTC comprises two key blocks: speed control and torque control.
The block diagram shows that DTC has two fundamental sections: the Torque Control Loop and the Speed Control Loop. Now we will walk around the blocks exploring each stage and showing how they integrate together.
with tachometers or encoders if the static speed accuracy requirement is over 0.5%, as it is for most industrial applications. This is a significant advance over all other AC drive technology. The Motor Model is, in fact, key to DTC’s unrivalled low speed performance. The Motor Model outputs control signals which directly represent actual motor torque and actual stator flux. Also shaft speed is calculated within the Motor Model. Torque comparator and flux comparator: The information to control power switches is produced in the Torque and Flux Comparator. Both actual torque and actual flux are fed to the comparators where they are compared, every 25 microseconds, to a torque and flux reference value. Torque and flux status signals are calculated using a two level hysteresis control method. These signals are then fed to the Optimum Pulse Selector. Optimum pulse selector: Within the Optimum Pulse Selector (ANN) to determine the switching logic of the inverter. Furthermore, all control signals are transmitted via optical links for high speed data transmission. This configuration brings immense processing speed such that every 25 microseconds the inverter’s semiconductor switching devices are supplied with an optimum pulse for reaching, or maintaining, an accurate motor torque. The correct switch combination is determined every control cycle. There is no predetermined switching pattern. DTC has been referred to as “just-in-time” switching, because, unlike traditional PWM drives where up to 30% of all switch changes are unnecessary, with DTC each and every switching is needed and used.
2.1. DTC’s Torque Control Loop VI measurements: In normal operation, two motor phase currents and the DC bus voltage are simply measured, together with the inverter’s switch positions. Adaptive motor mode: The measured information from the motor is fed to the Adaptive Motor Model. The sophistication of this Motor Model allows precise data about the motor to be calculated. Before operating the DTC drive, the Motor Model is fed information about the motor, which is collected during a motor identification run. This is called autotuning and data such as stator resistance, mutual inductance and saturation coefficients are determined along with the motor’s inertia. The identification of motor model parameters can be done without rotating the motor shaft. This makes it easy to apply DTC technology also in the retrofits. The extremely fine tuning of motor model is achieved when the identification run also includes running the motor shaft for some seconds.
2.2. Speed Control Torque reference controller: Within the Torque Reference Controller, the speed control output is limited by the torque limits and DC bus voltage. It also includes speed control for cases when an external torque signal is used. The internal torque reference from this block is fed to the Torque Comparator. Speed controller: The Speed Controller block consists both of a PID controller and an acceleration compensator. The external speed reference signal is compared to the actual speed produced in the Motor Model. The error signal is then fed to both the PID controller and the acceleration compensator. The output is the sum of outputs from both of them. Flux reference controller: An absolute value of stator flux can be given from the Flux Reference Controller to the Flux Comparator block. The ability to control and modify this absolute value provides an easy way to realize many inverter functions such as Flux Optimization and Flux Braking.
2.
DTC CONTROL THEORY
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
2.3. Algorithm for Optimizing ANN Algorithm is summarized as follows: 1. Make use of multi or bit-optimizing algorithm to learn all data samples, so as to obtain the weight values and threshold values of the first function network. 2. Filter all the samples. The filtering method is as follows: (a) Construct the BP network with the weight values and threshold values of the first function network. (b) Calculate the absolute value of difference which subtract the neural network outputs from the actual outputs. (c) Select the samples who’s the absolute value is bigger than the required accuracy as the new data sample of the second function network. At the same time, revise the original samples as follows: Make the actual outputs of the data included in the new samples be 1; make the actual outputs of the data included in the remaining samples be 0. Hereinafter such sample is named as “samples tagged by function network”.
961
The torque value obtained by dividing the turbine power by turbine speed is formed obtained as follows:
where Ct (λ) is the torque co-efficient of the turbine, given by
where the power co-efficient Cp is given by
where
5.
Figure 3.1. Block diagram of DTC based ANN for wind turbine driven induction generator set.
3. Train the neural network with the isolated new samples and the samples tagged by function network respectively. The resulting weight values and the threshold values from the training of the isolated new samples will be used in the elaborate function network, and the resulting weight values and the threshold values from the training of the samples tagged by function network will be used in the input identifier.
4.
Theoritical Wind Turbine Model
There are two types of wind turbines namely vertical axis and horizontal axis types. Horizontal axis wind turbines are preferred due to the advantages of ease in design and lesser cost particularly for higher power ratings [21]. The power captured by the wind turbine is obtained as
where the power coefficient Cp is a nonlinear function of wind velocity and blade pitch angle and is highly dependent on the constructive features and characteristics of the turbine. It is represented as a function of the tip speed ratio λ given by following equation. It is important to note that the aerodynamic efficiency is at the optimum tip speed ratio.
DTC Based on ANN Algorithm
The ANN is trained by a learning algorithm which performs the adaptation of weights of the network iteratively until the error between target vectors and the output of the ANN is less than an error goal. The most popular learning algorithm for multilayer networks is the back-propagation algorithm and its variants [l9]. The latter is implemented by many ANN software packages such as the neural network toolbox from MATLAB. In the case presented in this paper the DTC control strategy shown on table I have been implemented. Neural network has been devised having as inputs the torque error, the stator flux error and the position of the stator flux and as output the voltage space vector to be generate by the inverter [17]. The ANN is trained by a learning algorithm which performs the adaptation of weights of the network iteratively until the error between target vectors and the output of the ANN is less than an error goal. The most popular learning algorithm for multilayer networks is the back-propagation algorithm and its variants [l9]. The latter is implemented by many ANN software packages such as the neural network toolbox from MATLAB [19] [20]. In the case presented in this paper the DTC control strategy shown on table-I has been implemented. Neural network has been devised having as inputs the torque error, the stator flux error and the position of the stator flux and as output the voltage space vector to be generate by the inverter [17]. Stator flux and torque estimation: The components of the current (Isα, Isβ), and stator voltage (Vsα, Vsβ) are obtained by the application of the transformation given by (5) and (6), [1] :
The components of the stator flux (ϕsα, ϕsβ) given by (7).
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
The stator flux linkage phase is given by (8).
The electromagnetic couple be obtained starting from the estimated sizes of flux (ϕsα,ϕsβ and calculated sizes of the current, Isα Isβ)
The stator resistance Rs can be assumed constant during a large number of converter switching periods Te. The voltage vector applied to the induction motor remains also constant one period Te. Therefore, resolving first equation of system leads to:
962
function. The necessary steps to adjust these weights associated with the hidden neurons can be made through the training of the neurons. Levenberg- Marquardt back-propagation method is used here for training the network [18]. The inputs given are‘daxis stator flux’ and ‘q-axis stator flux’. Second neural network is used to determine the sector number for the estimated value of θe. There are total of six sectors, each sector of 60 degree. Again three layers of neurons are used but with a 5-4-1 feed forward configuration as shown in Fig. 6. Input layer is of log sigmoid transfer function, hidden layer is of hyperbolic tangent sigmoid function and the output layer is of linear transfer function. The training method used was Levenberg-Marquardt back-propagation. The input given is the angle theta since sector selection is purely based on theta. Last neural network is for the selection of voltage vector as given in Fig. 7, which is based on three inputs, flux, torque and the sector. Network taken this time is a 3-5-1 feed forward network with first layer of log sigmoid transfer function, second layer of hyperbolic tangent sigmoid transfer function and third layer of linear transfer function. Training method used was again Levenberg-Marquardt back-propagation. All the three neural networks were trained to performance 0.00001 mse. Here ‘mse’ is a network performance function and it measures the network’s performance according to the mean of squared errors (mse).
where φs0 stands for the initial stator flux condition. This equation shows that when the term RsIs can be neglected, (in high speed operating condition for example), the extremity of stator flux vector Vs. Furthermore, the instantaneous flux speed is only governed by voltage vector amplitude [1][4]. ANN based voltage vector estimator: Here we have used a feed forward neural network to select the voltage vector. For this purpose different configurations of networks were used and the best configured network is proposed and this scheme is depicted in the Fig. 2. The relation of variables used in the proposed scheme is as shown in Fig. 3.There are three neural networks. First is to estimate the value of stator flux position, θe as shown in the Fig. 5.
Figure 5.2. Schematic representation of three layered ANN deriving rotor angle from stator flux.
ANN speed controller: The input and output of the ANN controller can be obtained from the PI controller input output and which can be written as: (13)
where X(s) is the input and Y(s) is the output of PI controller, Kp and Ki are the proportional and integral gain constants.
The equation (13) can be written in the difference form as: Figure 5.1. Phasor diagram representation of stator flux, Clarks transformation components (α and β).
This is the angle between the stator flux and the rotor flux. It is a two input-one output feed-forward network with three layers. The input layer has 6 neurons of hyperbolic tangent sigmoid transfer function, first hidden layer has 4 neurons of log sigmoid transfer function and the output layer has 1 neuron of linear
(14)
where n is present time constant and (n-1) is previous time constant. The equation for speed controller can be obtained as:
(15)
Y (n) is the output of speed controller which is the controlling torque for the present control scheme of induction motor drive.
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
963
Figure 5. Speed to torque relation using ANN system for three layer system with weighing functions and relationship with arrow.
The ANN based speed controller (ANNSC) structure is as shown in Fig. 8.
6. Verification of the system using MATLAB/ SIMULINK The proposed system was implemented using MATLAB/ Simulink software; the wind turbine system is taken from matlab library. The complete system is a closed loop system with speed output of the machine compared with the reference. The ANN feed forward self learning technique is used to estimate the speed of the machine using current and flux parameters. The ANN will have flux and current as inputs, with speed estimation as output. The speed is controlled by using transfer function, which gives the desired torque for DTC calculation. The subsystem “DTC” mechanism is shown in the figure 6.1. The induction machine model was done referring to [1, 2, 3, 4, 8]. This machine model can work as a generator and also as motor with changes in torque value from negative to positive. Negative torque implies, the machine works as generator and positive torque for motor. But in this paper, we have considered generator characteristics, driven by wind turbine. The DTC implementation was [5,6, 9,10], of which [9] gives very detailed overview of industrial requirement, present challenges researchers facing etc. So, definitely a good technique for robust control is necessary. Papers [10-18] give an idea of DTC implementation, different techniques to improve system performance etc.
Figure 6.2. Design of complete DTC mechanism.
The space between the circles in figure 7.2 represents the hysteresis band in stator flux linkage amplitude. It is equal to the rated flux when operation below the base speed is called for.
Figure 6.3. Flux table.
6.1. DTC implementation: We will be getting the desired flux linkage from the speed reference using flux table. Stator flux can be controlled in both amplitude and angle by selectively applying proper voltage vectors. The required voltage-vectors for each segment of the stator flux position for increasing or decreasing torque and flux are indicated in each switching table.
From the fig 6.2, we can observe that the flux linkage can be calculated from speed reference. The DTC subsystem includes, table to convert estimated flux, stator flux, and electromagnetic torque, reference torque to get the three phase voltages Ua, Ub and Uc fig 6.1. The subsystem for table (fig 6.3) shows the construction of estimated parameters to get torque estimated. The PHIs is stator flux ( Isd, Isq) is converted from 2 to 3 dimensional by using inverse parks transformation to get the voltage parameter for logic table as shown in fig 6.4. The PHIs is stator flux (magnitude
Figure 6.1. Construction of the desired system using MATLAB/ SIMULINK.
Figure 6.4. Subsystem for table.
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
Figure 6.5. Torque parameters with stator flux comparison.
of Isd, Isq), of the machine output is compared with the reference flux estimated using flux table with a relay to get the phase voltages. When comparing estimated torque and actual torque using torque hysteresis shown in fig 6.5, we get the terminal voltage e-T. The parameters from fig 6.4, from PHIs we get stator voltage, e-PH and e-T, and the logic table and transport delay, we will be getting phase voltages Ua, Ub and Uc as shown in fig 6.1 and these variables will be input for the inverter system as shown in fig 6.6. Using this circuit we get the input voltage (both phase voltage and line voltage) to the wind turbine and induction generator set. The combinational logic for this inverter design is [0 0 0;-1/3 -1/3 2/3;-1/3 2/3 -1/3;-2/3 1/3 1/3; 2/3 -1/3 -1/3; 1/3 -2/3 1/3; 1/3 1/3 -2/3; 0 0 0]; The torque table parameters combinational logic [1 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0; 1 0 1; 0 0;0 0; 0 1 1;1 0; 0 0 0;0 0 0; 0 1 0; 0 1 1; 0 0 0;0 0 0; 1 1 0; 0 0 1; 0 0 0;0 0 0;1 0 0; 0 0 1; 0 0 0;0 0 0; 0 1 0; 1 0 1; 0 0 0;0 0 0;1 1 0; 1 1 0; 0 0 0;0 0 0; 1 0 1; 0 1 0; 0 0 0;0 0 0;0 0 1; 1 0 0; 0 0 0;0 0 0; 0 0 1; 1 1 0; 0 0 0;0 0 0;0 1 1; 0 1 0; 0 0 0;0 0 0; 1 0 0; 0 1 1; 0 0 0;0 0 0;1 0 1; 0 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0 0;0 0 0];
964
Figure 6.7. Subsystem for the design of inverter circuit for the wind turbine IG set.
given by
(16)
(17)
Figure 6.8. Implementation of induction machine model.
Mechanical sub-model of induction motor from the torque balance equations and neglecting viscous friction, the rotor speed ωo may be obtained as follows
Figure 6.6. Torque hysteresis control.
6.2. Induction generator implematation in Matlab/Simulink The phase voltage output from the inverter circuit is fed to the IG set, and the implementation is shown in fig 6.7. The torque input to this machine, we get from the wind turbine system. Electrical sub-model of the induction motor the threephase to two-axis voltage transformation is achieved using the following equation. Where Vas, Vbs, and Vcs are the three-phase stator voltages, while Vds and Vqs are the two-axis components of the stator voltage vector .torque sub-model of induction motor In the two-axis stator reference frame, the electromagnetic T is
(18)
where J is the moment of inertia of the rotor and load and TL is the load torque Stator current output sub-model. The stator current output sub-model is used to calculate the stator current amplitude according to the following equation
(19)
The machine parameters are as follows: Rs=2.7ohm, Rr=2.23ohm, Ls=0.3562H, Lm=0.3425 H, L r=0.3562 H, J=0.00825 kg m2 , P=2, Ts=0.0546 nm, Tr=0.160 nm.
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
965
6.3. Neural Network Implementation The figures below represent the response of electromagnetic torque, flux, stator current and the sequences state of switch inverter. The reference torque г*is a sample of Nm and a flux reference Ψs*= 1Wb. Reference model equations: The reference rotor flux components obtained from the reference model are given by:
reference torque. Purelin, transig are the linear transfer function blocks and bias is a constant with value “-2.316110024977692”.
Figure 6.11. Block diagram representation of Layer1 ANN system.
(20)
Adaptive model equations: The rotor flux components obtained from the adaptive model are given by:
(21)
Error betwen two model: Finally the adaptation scheme generates the value of the estimated speed to be used in such a way as to minimize the error between the reference and estimated fluxes. In the classical rotor flux MRAS scheme, this is performed by defining a speed tuning signal εω , to be minimized by a PI controller which generates the estimated speed which is fed back to the adaptive model. The expressions for the speed tuning signal and the estimated speed can be given as [13],[14]: Figure 6.12. MATLAB design of layers and weighing function.
(22)
The subsystems for the Neural Network block are shown below.
Figure 6.9. Subsystem for neural network.
The x{1}, y{1} are input, output for the system with two layers self learning weighed Neural Network. Fig 6.10 shows the process input subsystem, which is used to convert the given system with current and flux input to weighing function derivative. The block diagram representation of layer 1 of ANN function is shown in fig 6.10.
Figure 6.10. Process input subsystem.
The parameters with weighing function, dot-product and transformed function are shown in fig 6.11. Based on the weighing function and other respective parameters, we will get the first intermediate layer of the ANN system. As the system is bounded within the weighing function, the system will be stable and is self learning mechanism. Similarly, the weighing functions to the second layer also programmed to get the desired
Figure 6.13. Block diagram represented using MATLAB for the second layer of the system.
6.4. Wind turbine system modeling The block is taken from matlab library to implement the desired problem. The model is based on the steady-state power characteristics of the turbine. The stiffness of the drive train is infinite and the friction factor and the inertia of the turbine must be combined with those of the generator coupled to the turbine. The output power of the turbine is given by the following equation Pm= Cp(λ,β)
ν3wind,
where Pm is the mechanical output power of the turbine (W), cp the performance coefficient of the turbine, ρ the air density (kg/ m3), A the turbine swept area (m2), vwind the wind speed (m/s), λ the tip speed ratio of the rotor blade tip speed to wind speed, and β the blade pitch angle (deg). In this paper,there are two cases for the generator speed (1) constant generator speed of 500rpm, and (2) the generator speed is variable, we use a timer for this process. The pitch angle is taken as 22.5 degrees as constant and wind speed is variable with time and the wind speed magnitude varies with time as Time= [0 0.125 0.25 0.75 1.25 1.75 2.25 2.75 3.25 3.75]; Wind velocity= [0 8 12 15 20 10 14 6 18 14] against base speed of 10m/s.
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
7.
966
Simulation Results
Two case studies with constant generator speed and other variable generator speed. In both the cases, the wind speed changes as is described in section 6.4. 7.1. Case 1 Wind speed changes with place, time, season etc., for wind generation system, the wind rotates the blades, further it will rotate the wind shaft using some gear mechanism. The rotating shaft will be coupled to the induction generator or permanent magnet machine. We have considered induction generator in this study. Based on wind speed input to the wind turbine, electromagnetic torque will be the output. This torque is taken as reference in general for wind system In this case, we are considering constant generator speed, and examining with changes in the wind speed, how the induction generator torque output changes. For this we are considering the wind speed changes with time as shown in the fig 7.1. We have taken a timer block from MATLAB library and the timing and magnitude were as follows:
Figure 7.2. Graph between stator flux d and q axis.
Figure 7.3. Reference and actual value of electromagnetic torque.
Figure 7.1. Changes in wind speed (m/s) with time.
Examining the results, with changes in wind speed as in fig 7.1, the graph for stator flux d and q axis were shown in fig 7.2. Although with the changes because of wind speed, torque input changes to the generator, the trajectory between flux d and q does not change its circular motion. Hence the system is stable with transient response. Now considering the time of response and its deviation from reference value, Fig 7.3 and 7.4 will be explainable. Fig 7.3depicts, with reference torque change, the electromagnetic torque of the machine changes respectively. The deviation from reference to actual is as shown in Fig 7.4. Blue curve is reference value and green is actual value. The deviation is about 0.22seconds and error value is about ±0.03. The flux hysteresis band mainly affects the stator current distortion. Thus, for a fixed torque hysteresis band, the distortion increases with the flux hysteresis band. Some simulation results are shown for different values of the flux hysteresis band amplitude. Fig. 7.2 show the observable fact from the developed model. If small flux hysteresis band amplitude of ±0.01 Wb is applied, the stator flux vector locus approaches a circle and the
Figure 7.4. Zoomed picture of actual and reference electromagnetic torque.
With changes in torque input of the machine, the speed is maintained constant, which is considered as 500rpm in this case. The deviation from 500rpm is about +7.5rpm which can be negligible.
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
967
Figure 7.5. Generator speed with changes in electromagnetic torque.
Figure 7.6. Reference and actual speed comparison.
Figure 7.9. Phase voltage of IG for three phases.
Fig 7.4 shows the induction generator speed output. Fig 7.6 is helpful in examining the deviation of actual speed from reference speed as is self explainable. The three phase (a, b and c) currents are shown in the figure 7.7.
Figure 7.10. Zoomed view of three phase stator voltage.
The stator voltages for the 3phases are shown in Fig. 7.9, the voltages are also constant and doesn’t changed with torque change. The zoomed view of stator voltage is shown in Fig 7.10. Figure 7.7. Stator 3 phase current.
The stator current is constant as, even with the changes the torque, the speed of the machine is maintained constant, so current is also constant, the zoomed picture of the Fig 7.7 is shown in Fig 7.8.
7.2. Case 2 In this case, we are considering variable generator speed (comparing the constant speed in previous case), and examining with changes in the wind speed, how the induction generator torque output varies.
Figure 7.8. Zoomed picture of stator current 3 phases for Fig. 7.7.
Fig 7.11, 13 shows the trajectory of the estimated stator flux components DTC-SVM has as good dynamic response as the classical DTC.
Figure 7.11. Graph between stator flux d and q axis.
Also with the change in generator speed and torque, the
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
968
Figure 7.15. Variation of generator speed reference to actual.
Figure 7.12. Variation of induction generator speed with time.
actual machine speed follows without any delay and deviation. Hence DTC implementation using ANN system is very accurate and the neural network is self learning mechanism. With the changes in the wind generator speed, the stator current and voltage frequency is varied. When the induction generator reference speed varies at time 0.15, 1.5 and 3.75 seconds, the frequency of generator stator current and phase voltage changes as shown in Fig 7.16 and 7.17.
Figure 7.13. Reference and actual value of electromagnetic torque.
circular trajectory as in Fig 7.11 does not change. This trajectory graph is between stator flux d and q axis in pu. This shows that with transients in the form of speed variation or on load, the system will be in bounded state. As with change in wind speed, the electromagnetic torque varies. With respect to this input variation in torque (blue graph), the actual machine toque (green) followed. The deviations from actual value during the time (0 to 0.5seconds, 1.5 to 2seconds and 3.75 to 4 seconds) is due to the factor of variation of induction generator speed with time as shown in Fig 7.12. The deviation of actual and reference values are zoomed as shown in Fig 7.14. the deviation is nearly 0.05 seconds.
Figure 7.16. Stator three phase current.
Figure 7.17. Variation of stator phase voltages with time. Figure 7.14. Zoomed relation of reference (blue) and actual (green) torque parameters.
The generator reference speed can be observed from Fig. 7.12 and Fig. 7.15. With changes in the generator speed reference, the
As speed increases, the generator back emf increases and vice versa. Also with the back emf, the terminal voltage and frequency also increases. We can observe the changes in magnitude of voltage and frequency with changes in generator speed variation. The response of Electro Magnetic Torque (Fig 7.3 and 7.4)
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011
follows the reference values with a very small delay of 0.02seconds and also the ripples were minimum compared to classical DTC even at low torque. The simulation results prove that stator flux trajectory has never been violated from its elliptical path. The speed control of induction machine is totally in our hands from rated speed to one-third of its rated speed, and can be observed a stable operation (fig.7.14). From the above graphs, the proposed method will be very helpful for maintaining nearly constant voltage even with dynamic changes in load. With the electric power demand, we can supply the load centers with nearly constant voltage and frequency, but change in current. This proposed system will follow the reference path in less than 0.02seconds (fig 7.4, 7.15). Even the flux linkages were also maintained constant (fig 7.11), even with transient changes in load torque and also with machine speed. This technique holds good for normal Induction Motor Drive system, with positive torque coefficient.
8.
Conclusions
The performance has been tested by simulations. Also, a command flux optimization scheme has been proposed to reduce the torque ripple. The optimization was tested using simulation. The results show a reasonable improvement by flux optimization. The main improvements shown are: • Reduction of torque and current ripples in transient and steady state response. • No flux droppings caused by sector changes circular trajectory. • Fast stator flux response in transient state. In order to improve system performance, ANN is then used here. Induction Generator with wind turbine driver simulation model with intelligent ANN-DTC is created and analyzed based on MATLAB. Simulation results demonstrate the feasibility and validity of the proposed ANN-DTC system. Standing in vivid contrast against traditional DTC system, after ANN control and DTC technology is used, the intelligent system effectively accelerates speed and torque response, reduces torque and flux ripple, achieves fixed switch frequency and improves system performance. The proposed intelligent DTC simulation model provides an effective method for studying DTC. With respect to classical or schemes adopted before, this technique is enhanced in faster response to reference speed and torque, minimum rippled current, stator flux, maintaining constant voltage, can be adaptable as a generator for wind turbine or any prime-mover, or as a motor.
9.
Related Works
Lot of research work is being done to control the alternating current machines. Some control mechanisms like sensor less DTC, Extended Kalman Filter, SVPWM, and Artificial Intelligence based techniques with some recent trend, has led to a promising control of machine parameters, also improves efficiency of overall system. The work in this paper has got inspiration from “Design of self-tuning pi-type fuzzy controller [22], robust control of Induction Motor Drive (IMD) using SVM technique [23, 24]”, So studying these papers, and present problems in wind-turbine system, speed and power control, the proposed technique in this paper might be helpful.
Appendix The machine parameters are as follows: Rs=2.7ohm, Rr=2.23ohm, Ls=0.3562H, Lm=0.3425 H, L r=0.3562 H, J=0.00825 kg m2 , P=2, Ts=0.0546 nm, Tr=0.160 nm.
969
Acknowledgments The authors would like to thank Secretary, Mr. D.Dhanunjaya, Principal, Dr. K.Prabhakar, HOD, Sri M.Swetha Chalapathy, all teaching and non teaching staff of VITAM college of Engineering, Visakhapatnam, India for providing us good research environment, assisting and encouraging us in many forms, without them this work may not successful. I am very much grateful to Editor in Chief of IJRRCS©: Science Academy Publisher United Kingdom, Mr. Jonathan Loo, and Editorial Board Members, reviewers for accepting and publishing this in this prestigious Journal. Author 1 would like to thanks Mr. M.Avinash, Dr. A. Santhi Swaroop, Mr.A. Anada Swaroop, Mr. D.V.S.Sekhar for constant encouragement towards research and final reading in making this document and spellings checking, last but not the least, thanks to our parents, family and friends for everything.
References [1] Krause, P. C., ‘Simulation of symmetrical induction machinery’, IEEE Trans. Power Apparatus Systems, Vol. PAS-84, No. 11, pp. 1038–1053 (1965) [2] Ghani, S. N., ‘Digital computer simulation of three-phase induction machine dynamics — a generalized approach’, IEEE Trans Industry Appl., Vol. 24, No. 1, pp. 106–114 (1988) [3] Wade, S., Dunnigan, M. W. and Williams, B. W., ‘Modeling and simulation of induction machine vector control and rotor resistance identification’, IEEE T rans. Power Electronics, Vol. 12, No. 3, pp. 495–505 (1997)172 [4] Shi, K. L., Chan, T. F. and Wong, Y. K., ‘Modeling of the three-phase induction motor using SIMULINK’, Record of the 1997 IEEE International Electric Machines and Drives Conference, USA, pp. WB3-6 (1997) [5] Shi, K. L., Chan, T. F. and Wong, Y. K., ‘Modeling and simulation of direct self control system’, IAST ED International Conference: Modeling and Simulation, Pittsburgh, USA, pp. 231–235 (May 1998) [6] Trzynadlowski, A. M., T he Field Orientation Principle in Control of Induction Motors, Kluwer (1994) [7] Using SIMUL INK, Dynamic System Simulation for MATLAB, The Mathworks Inc. (1997) [8] Krause, P. C., Wasynczuk, O. and SudhoV, S. D., Analysis of Electric Machinery, IEEE (1995) [9] Technical Guide No.1- Direct Torque Control http://www.abb.com/ motors&drives [10] Kuo-Kai Shyu, Li-Jen Shang, Hwang-Zhi Chen and Ko-Wen Jwo: Flux Compensated Direct Torque Control of Induction Motor Drives for Low Speed Operation. IEEE Transactions on Power Electronics, vol. 19, no. 6, November 2004. [11] K. L. Shi, T. F. Chan, Y. K. Wong, S. L. Ho: Estimation of an Induction Motor Drive Using an Optimized Extended Kalman Filter. IEEE Transaction on Industrial Electronics, vol. 49, no. 1, February 2002. [12] T. Brahmananda Reddy, B. Kalyan Reddy, J. Amarnath, D. Subba Rayudu, and Md. Haseeb Khan: Sensorless Direct Torque Control of Induction Motor based on Hybrid Space Vector Pulsewidth Modulation to Reduce Ripples and Switching Losses – A Variable Structure Controller Approach. IEEE Power India Conference, 10-12 April 2006. [13] Giuseppe S. Buja, Marian P. Kazmierkowski: Direct Torque Control of PWM Inverter-Fed AC Motors – A survey. IEEE Transactions on Industrial Applications, vol. 51, pp. 744–757, August 2004. [14] M. Depenrock, “Direct self-control of inverter-fed machine,” IEEE Trans. Power Electron., vol. 3, pp. 420-429, Oct. 1988. [15] I. K. Buehring and L.L. Freris, “Control policies for wind energy conversion systems”, Proc. Inst. Elect. Eng. C, vol. 128, pp. 253-261, Sept. 1981. [16] Mukund R. Patel, Wind and Solar Power Systems, CRC Press, 1999. [17] S. N. Bhadra, D. Kastha, S. Banerjee, Wind Electrical Systems, OXFORD University Press, 2005. [18] J.J. Brey, Castro, A., Moreno, E., and Garcia, C., “Integration of renewable energy sources as an optimised solution for distributed generation,” IECON 02 [Industrial Electronics Society, IEEE 2002 28th Annual Conference of the, Vol. 4, 5-8 Nov., pp.:3355 – 3359, 2002. [19] S.H.Chichester, Grid integration of wind energy conversion systems, Wiley, 385p, 1999. [20] R.Toufouti S.Meziane ,H. Benalla, “Direct torque control for induction motor using intelligent techniques” , Journal of Theoretical and Applied Information Technology, pp 35-44, 2007. [21] Kelvin Tan,Syed Islam:” Optimum control strategies in energy conversion of pmsg wind turbine system without mechanical sensors”,IEEE Transac-
D.V.N. Ananth et al. / IJRRCS, Vol. 2, No. 4, pp. 959-970, August 2011 tions on Energy Conversion, vol. 19, no. 2, June 2004, pp.392-399 [22] Jose L. Azcue P. and Ernesto Ruppert, “Three-phase Induction Motor DTC-SVM Scheme with Self-Tuning PI-Type Fuzzy Controller” IEEE 2010 Seventh International Conference on Fuzzy Systems and Knowledge Discovery (FSKD 2010) , June 2010, pp. 757-762. [23] SHANG De-xi,LIU Yan-cheng,SUN Fan-jin,ZHANG Hai-yan, “Study on DTC-SVM of PMSM Based on Propeller Load Characteristic” IEEE World Congress on Intelligent Control and Automation, June 25 - 27, 2008, pp. 6445-6449. [24] Farouk M. Abdel-kader, A. EL-Saadawi, A. E. KALAS, Osama M.EL-baksawi, “Study In Direct Torque Control of Induction Motor By Using Space Vector Modulation” , Aug 2008, pp-224-229. [25] Chengzhi Cao, Fengkun Li, Kun Zhang, Hongbing Zhang, Hongli San, “Chaos Optimizing BP-NNG Speed Recognition in DTC System”, IEEE International Symposium on Intelligent Information Technology Application Workshops, Aug2008, pp. 1077-1080.
970