CHAPTER-5 DESIGN OF DIRECT TORQUE CONTROLLED INDUCTION MOTOR DRIVE

Transcription

1 113 CHAPTER-5 DESIGN OF DIRECT TORQUE CONTROLLED INDUCTION MOTOR DRIVE 5.1 INTRODUCTION This chapter describes hardware design and implementation of direct torque controlled induction motor drive with high-speed digital signal processor. Implementation of different control algorithms necessarily requires hardware design and implementation. In this research, prototype hardware is developed with digital signal processor to verify the control strategies. In order to validate the developed drive, the experimental results are obtained for different direct torque control strategies including the proposed strategy. 5.2 EXPERIMENTAL DRIVE SYSTEM The experimental drive system for a 3-phase squirrel cage induction motor comprises of three subsystems. 1. Power circuit 2. Control circuit 3. Sensing circuitry and signal conditioning circuitry

2 114 In this chapter, these three subsystems are explained with the block diagram and hardware of experimental setup in detail. Block diagram of experimental setup is shown in Figure Power circuit Power circuit consists of an uncontrolled rectifier, filter and IGBT based voltage source inverter. These are explained in the following sections in detail Voltage Source Inverter Voltage source inverter takes a DC bus voltage and uses six switches arranged in three phase legs as shown in Figure 5.2. From the middle of each phase leg comes from the line, which connects the stator of the motor. The voltage on these lines must be a balanced three-phase sinusoidal waveform in order to drive the induction motor. This is achieved by a controlled switching to the gate of the IGBTs. Bridge Rectifier Filter Voltage Source Inverter Induction Motor Drive Circuit Voltage and Current Sensing Circuit TMS320LF2407 Signal Conditioning circuit Figure 5.1 Block diagram of Experimental Set-up

3 115 1MBH10D-060 1MBH10D-060 1MBH10D-060 VDC 1MBH10D-060 1MBH10D-060 1MBH10D-060 Phase B Phase A Phase C Figure 5.2 Power circuit diagram of Voltage source inverter Selection of Power components Nowadays, Power Components used in industrial motor drives are MOSFETs and IGBTs. IGBT switches are used in inverters for drives applications. They are replacing MOSFETs in many high voltage, hard switching applications since they have lower conduction and switching losses for the same output power. They are lower cost devices and have smaller input capacitance. Most IGBT modules are used in hard switching applications of up to 20 khz beyond which, switching losses become very significant. Since MOSFETs have these drawbacks, in this experimental work, IGBTs are preferred in inverter circuit. Table 5.1 shows the switching performance and characteristics of IGBT used in power inverter circuit. Turn-on time of IGBT used in inverter is 1.2 microseconds and rise time is 0.6 microseconds. The typical fall time is 200 nano seconds.

4 116 Table 5.1 Characteristics of IGBT Characteristics Values Device IGBT Type 1MBH10D-060 Current Rating 10A Voltage Rating 600V R ON at T j =25 C 0.23 R ON at T j =150 C 0.22 Fall Time (typical) 200 nsecs Drive Type Voltage Drive Power Minimum Drive Complexity Simple Current Density for a given voltage Drop High Switching Losses Low Snubber circuits Snubbers are needed to protect the switches (IGBT) against over voltage transients resulting from current changing due to the parasitic inductance. In addition to providing protection from over voltage, snubbers can be employed to: Limit di (or) dt dv dt Shape the load line to keep it within the safe operating area (SOA) Transfer of power dissipation from the switch to a resistor Reduce total switching losses Reduce voltage and current ripples

5 117 Figure 5.3 RCD snubber circuit RCD snubbers are used in this research work to protect the IGBT inverter as shown in Figure 5.3. Direct mount snubbers are used in which hyper fast, soft recovery diode MUR3060 was prefered. Direct mount types have lower inductance due to flat, radial lead geometry. They are installed by using their own screws. Direct mount capacitors are rated for higher current because heavy copper lugs are connected directly to the capacitor element. SCD polypropylene double metalized capacitance is used in this snubber circuit Selection of heat sink The selection of a heat sink is constrained by many factors including set space, actual operating power dissipation, heat-sink cost, flow condition around a heat sink and assembly location. Table 5.2 provides a comparison of the percentage transfer efficiency of the different type of heat sinks for natural airflow conditions. It is understood from the table that Ducted Pin Fin, Boded &Folded Fins have

6 118 got higher percentage transfer efficiency. In this research, Ducted Pin Fin heat sink was used. Table 5.2 Heat Sink Type vs. Percentage Transfer Efficiency Heat Sink Type %Transfer Efficiency Stampings &Flat Plates Fined Extrusions Impingement Flow Fan Heat Sinks Fully Ducted Extrusions Ducted Pin Fin, Bonded &Folded Fins the power inverter All the above points are taken into consideration while designing Control Circuits The DTC control algorithm is performed by utilizing a DSP controller board ezdspf2407. The optimal switching patterns, which are selected based on the flux and torque status, are stored in a look-up table Digital signal processor (TMS320LF2407) The ezdspf2407 board is available from Texas instruments as a development tool, is shown in Figure 5.4. It is one of the processors that execute most of the programs of the control algorithm. The DSP kit provides a complete development environment, and includes the DSP board, power supply, on-board JTAG compliant emulator and specific version of the Code Composer Studio Integrated development Environment. The DSP board itself

7 119 has nearly all peripheral signals available on the board headers, making it easy to interface the board with other system hardware. Figure 5.4 TMS320LF2407 ezdsp DSP Starter Kit Some of the hardware features are given below: Clock frequency is 40 MHz and 30 MIPS Communication interface for SCI (serial communication interface) and CAN (control area network) controller 10 bit ADC with twin auto sequencer Serial components connected to SPI (serial peripheral interface) such as serial DAC, serial EEPROM and serial LED driver Serial communication interface module

8 analog channel and two event manager for PWM generation (both asymmetrical and symmetrical PWMs) Voltage and Current sensors that interface to the capture and quadrature encoded- pulse (QEP) decoding logic. Capture pins to capture the logical signals. External Memory interface Gate driver and opto-isolator: The gate drivers are used to get the signal pulses from the control board and amplify them to the level required for switching the IGBTs and opto isolators are used to isolate the power and control circuits Sensing and signal conditioning circuitry Hall Effect sensors are used as voltage and current sensors. Two voltage sensors and two current sensors are used. Apart from the sensing circuitry, signal-conditioning circuits are also designed to meet the requirements of DSP Voltage sensing circuit Hall Effect voltage sensors are widely used to measure the phase voltages (LV-25-P-29). The voltage obtained from the voltage sensor is bipolar voltage. But, DSP can accept only unipolar voltage. In order to get the unipolar voltage, a unipolar converter circuit is designed and implemented. Once the processor receives the input, the received value is scaled in the control algorithm.

9 Current Sensing Circuit Current sensors are used to obtain the proper information of current from the phase windings of induction motor. In this research work, two closed loop current sensors of PC board mount type LA-100P are used. The output of the current sensor for the given input current is (0-3) V Isolation circuits If the voltage and current sensor output are within the 0 to 3V input range of the ADC, with no significant noise, and meets the source impedance requirement of the ADC then a direct connection between the sensor output and ADC input is possible. But in most cases, operational amplifier is used to meet the input impedance requirement of the ADC and also used as isolation between sensor and processor control unit. 5.3 EXPERIMENTAL PROCEDURES FOR DIFFERENT DTC STRATEGIES + - RAM Device Firing Inverter PC DSP Data and address bus 10-bit ADC Phase Voltages IM Phase Currents Figure 5.5 DSP based control system for DTC

10 122 The overall DTC layout is implemented on the 40 MHz TI TMS320LF2407 DSP based drive system as shown in Figure 5.5. This DSP kit provides a complete development environment and includes the main DSP board, power supply for the board, on board JTAG compliant emulator and an ezdsp specific version of the code composer studio. The algorithm was programmed in C. An efficient C code optimizer is employed during compilation. 150 khz bandwidth Hall-effect sensors are used to measure the phase currents. It is sufficient to measure two phase currents only. The ezdsp captured the resultant signals with 10 bit flash analog to digital converters (ADC) at the start of every interrupt. The ADCs are triggered in hardware by the DSP internal interrupt clock. The phase voltages are also measured by using Hall Effect voltage sensors. The DSP communicates with an input card and fires the inverter devices. (i) (ii) (iii) Control Update Period: The time interval between successive calculations and generation of new voltage vector demands. Control update period in this research work is 50μsec. Inverter switching period: The minimum time interval between changes in the output voltage vector state. A change in the output voltage vector can result from any one signal leg switching, so it is possible for the inverter switching period to be less than the leg switching period. Inverter switching period is also 50μsec. Device switching period: The minimum time interval during which an individual power device in the six device bridge is allowed to switch ON and OFF. Because the bridge consists of three legs, each containing two devices switched

11 123 in opposition, the device switching period is equal to the leg switching period and the value is 100μsec. In this research work, the control update and inverter switching periods are identical and equal to half the device switching period. The various time periods are limited either by the switching devices or by the processor. Figure 5.6 shows the photographical view of complete experimental set-up in which, power inverter, induction machine coupled with a d.c. generator, DSP, oscilloscope and personal computer are shown. D.C. generator with resistive load is also shown in this Figure. modes: The experimental control system is made and done in the following (i) (ii) Torque and Flux loops without speed encoder (Speed Estimation) Torque, Flux and speed loops (Online speed measurement)

12 Figure 5.6 Photographical view of Experimental Set-up 124

13 Conventional DTC The motor phase currents and voltages are measured and the measured information is given to the motor model. Initial information of the motor is given to the model based on the principle of auto tuning. Using speed encoder the parameters of the motor model are determined. The output signals from the motor model represent the torque and stator flux directly and the speed is also estimated directly from the motor model. Power IGBTs are controlled by the information from the torque and flux comparators. Actual torque and flux values are compared with the respective reference values in the comparators for every 100 micro seconds. Then the torque and flux error signals are fed to the optimum switching state selector which is called as pulse selector. DSP TMS320LF207 is within the state selector to determine the switching state of the inverter. For every 100 microseconds, IGBTs are supplied pulses for maintaining the torque and flux in the required levels. To get the desired results torque and flux controller must be designed accurately. The torque controller gain was chosen a high value to achieve fast torque response. The control systems parameters chosen for this research are given below: Proportional Gain for flux comparator = K P = 100 Integral Gain for flux comparator = K I = 300 Proportional Gain for torque comparator = K PT = 2 Integral Gain for torque comparator = K IT = 200 The feedback rate of the speed loop = 1 khz. The speed loop bandwidth = [0, 32 Hz]. All graphics of the present chapter correspond to the experimental results obtained from a three phase induction motor. In all cases, the reference values are

14 126 Torque reference value = 1.0Nm. Torque hysteresis value = 0.27Nm. Flux reference value = 0.8Wb. Flux hysteresis value =0.035Wb. Sample time = Ts = 100μs. The DSP has been programmed using C language and assembler. All the tasks, whose execution time is critical, have been programmed in assembler. To implement the system, the first step taken for consideration is the delay due to non-ideal behavior of the whole system. The most significant delay is introduced by the ADC and the control algorithm. The control algorithms implemented in this research have three routines. They are given as follows: (i) (ii) (iii) ADC processing with all the data Estimation of torque and flux values Implementation of DTC The time taken to execute these routines is about 40 μs. The sampling frequency of ADC is set to 10 khz and therefore the sampling time is 100 μs. Totally, there is a delay of 140 μs to send the new VSI state from the sampled data. At every sampling time the voltage vector selection block chooses the inverter switching state that reduces the instantaneous flux and torque errors. Figure 5.7 shows the torque developed at 1000 rpm, using conventional DTC strategy. From this figure it is observed that the ripple

15 127 content are more and it is matched with the simulation results as shown in Chapter-3. As described in previous chapters the reference or command torque value is taken as 1 Nm and the steady state torque is running around the command torque in narrow band manner. Command torque is given from the torque reference controller which includes speed control loop also. The torque reference output is taken from this controller through Digital to Analog converters and given to the torque comparator. Figure 5.8 shows the stator flux linkage in stationary reference frame using conventional DTC strategy. Figures 5.9 and 5.10 show the stator flux linkages in d-axis and q- axis at stator with respect to stationary reference frames respectively. The flux linkages and torque have been estimated within the sampling period of 100μs. This is accomplished by sensing the stator current at the same sampling period. Figure 5.11 shows the output voltage across the PWM inverter terminals looking through attenuation probe. Figure 5.12 shows the position of stator flux in angle. Figures 5.13 and Figure 5.14 show the direct and quadrature currents and voltages respectively. Figure 5.7 Torque developed in induction motor using conventional DTC controller at 1000 r.p.m.

16 128 Figure 5.8 Stator flux in conventional DTC scheme Figure 5.9 Stator flux in stationary reference frame in DTC scheme (d-axis)

17 129 Figure 5.10 Stator flux in stationary reference frame in DTC scheme (q-axis) Figure 5.11 Output phase voltage across the inverter terminals through attenuation probe

18 130 Figure 5.12 Position of stator flux (angle) shown on computer screen Figure 5.13 Direct and Quadrature axes currents

19 131 Figure 5.14 Direct and Quadrature axes Voltages Intelligent DTC experimentally: The following Intelligent control algorithms are implemented (i) (ii) (iii) (iv) Fuzzy based Direct Torque control Neural Network based Direct Torque control Neuro Fuzzy based Direct Torque control Genetic algorithm based Direct Torque Control To implement intelligent control algorithms, the timing should be as precise as possible. The sampling time taken to execute intelligent control is also 100 μs. However, execution of intelligent direct torque control algorithm differs from classical DTC algorithm due to the intelligent control blocks presented such as Fuzzy, training of Neural Networks along with the DTC algorithm. In this algorithm, once the active state is sent then only the

20 132 Fuzzy logic or neural network algorithm starts being executed. Because, some part of this controller is executed in personal computer itself and the processor waits until execution is finished to obtain the duty cycle. Once the duty cycle is obtained, the timer is programmed. The time taken for this process is 100 μs. The duty cycle, which needs to change the active state before this time is ignored, and it is not taken for consideration. The total time taken to execute classical and intelligent control algorithms is as given in Table 5.2. Figure 5.15 (a) shows the pie chart for time taken for implementation of classical DTC, Figure 5.15 (b) shows the pie chart for time taken for implementation of Intelligent controlled DTC. It is noted that the time taken for all the intelligent control techniques is limited to be nearing the same. Figure 5.15 (c) shows the pie chart for time taken for implementation of proposed DTC strategy. Table 5.2 Timing for Various DTC strategies Sl.No. Control Algorithms Sample Time (μs) Delay Time for software Execution (μs) ADC interruption Time (μs) 1. Classical DTC Direct Torque Fuzzy logic Control Direct Torque Neural Network Control 4. Direct Torque Neuro Fuzzy Control 5. Genetic algorithm based direct torque control

21 us ADC interruption Time Sample Time Delay Time 200us 40us Figure 5.15 (a) Time taken for classical DTC ADC Interruption Time Sample Time Delay Time 100us 200us 80us Figure 5.15 (b) Time Taken for DTC using intelligent control techniques

22 134 ADC interruption Time Sampling Time 100us 200us Delay time 60us Figure 5.15 (c) Time Taken for proposed DTC strategy Figure 5.16 shows the torque developed using neural network based direct torque controller. Figure 5.17 shows the torque developed using Fuzzy based direct torque controller. Figure 5.18 shows the torque developed using adaptive neuro fuzzy based direct torque controller in the induction motor at 1000 rpm. Figure 5.19 shows the torque developed using genetic algorithm based direct torque controller at 1000 rpm in which neural networks are trained using genetic algorithm with binary coding representation. Figure 5.20 shows the torque developed using genetic algorithm based direct torque controller at 1000 rpm in which neural networks are trained using genetic algorithm with floating point coding representation.

23 135 Figure 5.16 Torque developed in induction motor using neural network based direct torque control at 1000 r.p.m. Figure 5.17 Torque developed in induction motor using Fuzzy logic based direct torque control at 1000 r.p.m.

24 136 Figure 5.18 Torque developed in induction motor using adaptive neuro fuzzy based direct torque control at 1000 r.p.m. Figure 5.19 Torque developed using Direct Torque Neuro controller trained by genetic algorithm (Binary coding representation)

25 137 Figure 5.20 Torque developed using Direct Torque Neuro controller trained by genetic algorithm (Floating Point representation) Proposed DTC The proposed DTC strategy is implemented through the same hardware setup, which has been used for other control strategies. As discussed in chapter-4, in the experimental implementation also, the following steps are involved: (i) (ii) (iii) (iv) (v) Calculation of torque and flux increments Calculation of stator flux adjacent angle Control of angle Δθ to define the position of new stator flux vector Calculation of stator flux vector increment Calculation of stator reference voltage

26 138 The sampling frequency and time taken for the execution of proposed strategy are 10 khz and 40μsec respectively. The delay time taken for execution of this control strategy is 60 μs, it includes the time taken for calculation of stator flux increment, torque increment and stator flux reference voltage. Figure 5.21 shows the torque developed at the application of proposed DTC strategy with the constant switching frequency and deadbeat strategy. In this figure the developed torque is with reduced ripple at 1200 r.p.m. It is noticed that the torque ripple is minimized when compare to classical DTC strategy and this Figure is followed by the torque developed using proposed strategy at the speed of 1000 rpm as shown in Figure Figure 5.23 shows the torque developed at the speed of 600 rpm using proposed constant frequency strategy In this figure it is observed that the ripple content is almost same as high speed operation Flux linkages are sinusoidal quantities in stationary reference frame and d.c. quantities in synchronously rotating frame. For the proposed DTC strategy, stator flux linkages are considered in synchronously rotating frame. Figure 5.24, 5.25 and 5.26 show the stator flux magnitudes at the different speeds such as 1200, 1000 and 600 rpm respectively. Figure 5.27 shows the locus of stator flux with reduced ripple in DTC scheme using proposed algorithm at 1000 r.p.m. It is noticed that the stator flux follows a smooth circular path. Figure 5.28 shows the stator flux angular advancement at 1000 rpm. These values are given in radians. Figure 5.29 shows the three phase stator currents of Induction motor at 1000 r.p.m at the application of constant switching frequency and deadbeat DTC strategy. Actual speed response of the motor is shown in Figure Speed reference taken here is 600 rpm. For 750 rpm the speed response is shown in Figure 5.30 and this speed response is captured on the computer screen.

27 139 Figure 5.21 Torque developed at the application of proposed algorithm at 1200 r.p.m Figure 5.22 Torque developed at the application of proposed algorithm at 1000 r.p.m

28 140 Figure 5.23 Torque developed at the application of proposed algorithm at 600 r.p.m Figure 5.24 Stator flux plot at the application of proposed algorithm of torque and flux ripple minimization at 1200 rpm

29 141 Figure 5.25 Stator flux plot at the application of proposed algorithm of torque and flux ripple minimization at 1000 rpm Figure 5.26 Stator flux plot at the application of proposed algorithm of torque and flux ripple minimization at 600 rpm

30 142 Figure 5.27 Locus of stator flux with reduced ripple in DTC scheme using proposed algorithm at 1000 r.p.m. Figure 5.28 Stator flux vector angular advancement

31 143 Figure Phase stator currents of Induction motor at 600 r.p.m. Figure 5.30 Speed Response at the application of proposed DTC strategy

32 144 Figure 5.31 Speed response on computer screen 5.4 CONCLUSION This chapter describes the design and implementation of direct torque control of induction motor system with clear design procedure and logic of different DTC strategies. A 1HP 3-phase squirrel cage induction motor is used for experimental study and ezdsp TMS320LF2407 is used to implement the control algorithms. The programs are written in C language and assembler language. The waveforms are captured by Code composer studio software package. In this chapter, the experimental wave forms of stator current, voltage, torque, speed, stator flux, position of stator flux for different control strategies are shown clearly. In experimental platform, all of the DTC strategies including the proposed DTC strategy were implemented and waveforms were captured. Moreover, to establish the benefit of proposed strategy, all the waveforms includes speed, torque and flux were captured. From the waveforms, it is observed that the torque ripple produced by the application of the proposed strategy is comparatively low. It is also observed that, the experimental results match with the simulation results.