Academia.eduAcademia.edu

Figure 10 – uploaded by Satyanarayana Mummana

See full PDFdownloadDownload figure
Indonesian J Elec Eng & Comp Sci Figure 10. Stator current THD based on a neural network for the PMSM DTC

Figure 10 Indonesian J Elec Eng & Comp Sci Figure 10. Stator current THD based on a neural network for the PMSM DTC

Related Figures (15)

Indonesian J Elec Eng & Comp Sci, Vol. 32, No. 2, November 2023: 676-687 The comparators are used to compare the calculated and reference values of the stator flux linkage and electromagnetic torque. There is a three-level torque comparator as well as a two-level stator flux comparator. As shown in Table 1, the estimated stator flux space vector sector in the complex plane and the comparator outputs serve as the inputs for the best switching lookup table.
Indonesian J Elec Eng & Comp Sci, Vol. 32, No. 2, November 2023: 676-687 The comparators are used to compare the calculated and reference values of the stator flux linkage and electromagnetic torque. There is a three-level torque comparator as well as a two-level stator flux comparator. As shown in Table 1, the estimated stator flux space vector sector in the complex plane and the comparator outputs serve as the inputs for the best switching lookup table.
Figure 1. PMSM's fundamental direct torque control system Figure | depicts the block diagram for the fundamental direct torque control approach for PMSM When little time period AWs t= V2 At? is applied to the voltage of stator, then, throughout this time, the stator flux space vector moves in the same direction as the applied stator voltage space vector. If the stator resistance (Rs) voltage drop is not included, the stator voltage equation can be expressed in a fixed reference frame as shown in (10):
Figure 1. PMSM's fundamental direct torque control system Figure | depicts the block diagram for the fundamental direct torque control approach for PMSM When little time period AWs t= V2 At? is applied to the voltage of stator, then, throughout this time, the stator flux space vector moves in the same direction as the applied stator voltage space vector. If the stator resistance (Rs) voltage drop is not included, the stator voltage equation can be expressed in a fixed reference frame as shown in (10):
Inputs for the direct torque control based PMSM model with flux error and change in flux error are shown in Figure 4 of the generated Simulink fuzzy controller model. A two-input Mamdani fuzzy inference system (FIS) with two rules is implemented via the gauss membership function. The knowledge of the users is used to create a set of language rules that are employed for control techniques, as illustrated in Table 2.
Inputs for the direct torque control based PMSM model with flux error and change in flux error are shown in Figure 4 of the generated Simulink fuzzy controller model. A two-input Mamdani fuzzy inference system (FIS) with two rules is implemented via the gauss membership function. The knowledge of the users is used to create a set of language rules that are employed for control techniques, as illustrated in Table 2.
. IMPLEMENTATION OF NEURAL NETWORK CONTROLLER IN DIRECT TORQUE CONTROL OF PMSM The fundamental DTC scheme of the PMSM is the foundation upon which the Simulink design for the neural network controller-based DTC of the PMSM is built. The block diagram for synchronous motor direct torque control using neural networks is shown in Figure 5. Because it has at least one feedback loop, a recurrent neural network, like in Figure 6, sets itself apart from other neural networks. One of the most used strategies for training a network is back propagation. During the training phase, the weights are initially initialized at random inside the network. The network's output is then collected and contrasted with the target value. Output layer's weights are adjusted based on the network error, which is estimated and applied. The weights of the earlier levels are updated similarly by propagating the network mistake backward.
. IMPLEMENTATION OF NEURAL NETWORK CONTROLLER IN DIRECT TORQUE CONTROL OF PMSM The fundamental DTC scheme of the PMSM is the foundation upon which the Simulink design for the neural network controller-based DTC of the PMSM is built. The block diagram for synchronous motor direct torque control using neural networks is shown in Figure 5. Because it has at least one feedback loop, a recurrent neural network, like in Figure 6, sets itself apart from other neural networks. One of the most used strategies for training a network is back propagation. During the training phase, the weights are initially initialized at random inside the network. The network's output is then collected and contrasted with the target value. Output layer's weights are adjusted based on the network error, which is estimated and applied. The weights of the earlier levels are updated similarly by propagating the network mistake backward.
Figure 6 depicts the suggested recurrent neural network's configuration, which consists of three layers: input layer, hidden layer, and output layer. The neurons are represented by the circles in the neural network.
Figure 6 depicts the suggested recurrent neural network's configuration, which consists of three layers: input layer, hidden layer, and output layer. The neurons are represented by the circles in the neural network.
Figure 6. RNN architecture-based neuro- controller ~ The input is linked at node X. The weights between the input layer and the hidden layer are made up of V11 and V12. The hidden layer is linked to the bias value, bias value connected to hidden layer (Bh). The weight values for the output layer are W11 and W21, and the bias value is bias value connected to output layer (Bo). The RNN is subjected to the back propagation training technique. The network's architecture weights are modified as:
Figure 6. RNN architecture-based neuro- controller ~ The input is linked at node X. The weights between the input layer and the hidden layer are made up of V11 and V12. The hidden layer is linked to the bias value, bias value connected to hidden layer (Bh). The weight values for the output layer are W11 and W21, and the bias value is bias value connected to output layer (Bo). The RNN is subjected to the back propagation training technique. The network's architecture weights are modified as:
Direct Torque Control of Synchronous Motor using Space Vector Modulation
Direct Torque Control of Synchronous Motor using Space Vector Modulation
Figure 8. Stator current total harmonic distortion for basic DTC scheme of PMSM Waveforms of the current for stator and ripple for torque generated using the PMSM's fundamental DTC approach is shown in Figure 11. The waveforms of the current for the stator and ripple for the torque obtained by utilising a fuzzy controller in the DTC of the PMSM are shown in Figure 12. The torque ripple and stator current waveforms generated by integrating neural network controllers into the PMSM's DTC are shown in Figure 13. In Table 3, the fundamental DTC of the PMSM and the stator current THD derived using fuzzy and neural controllers are contrasted with the DTC-based PMSM. From Table 3, it can be inferred that, in comparison to the basic DTC technique of PMSM, the THD values of stator current and steady-state torque ripple have decreased as a result of the inclusion of proposed neural network and fuzzy controllers in a loop of the DTC scheme of PMSM.
Figure 8. Stator current total harmonic distortion for basic DTC scheme of PMSM Waveforms of the current for stator and ripple for torque generated using the PMSM's fundamental DTC approach is shown in Figure 11. The waveforms of the current for the stator and ripple for the torque obtained by utilising a fuzzy controller in the DTC of the PMSM are shown in Figure 12. The torque ripple and stator current waveforms generated by integrating neural network controllers into the PMSM's DTC are shown in Figure 13. In Table 3, the fundamental DTC of the PMSM and the stator current THD derived using fuzzy and neural controllers are contrasted with the DTC-based PMSM. From Table 3, it can be inferred that, in comparison to the basic DTC technique of PMSM, the THD values of stator current and steady-state torque ripple have decreased as a result of the inclusion of proposed neural network and fuzzy controllers in a loop of the DTC scheme of PMSM.
Figure 9. Stator current THD-based DTC of the PMSM using fuzzy logic
Figure 9. Stator current THD-based DTC of the PMSM using fuzzy logic
Figure 11. Typical PMSM DTC configuration: current for stator and ripple for torque
Figure 11. Typical PMSM DTC configuration: current for stator and ripple for torque
Figure 12. PMSM DTC with fuzzy logic for current for stator and ripple for torque
Figure 12. PMSM DTC with fuzzy logic for current for stator and ripple for torque
Table 3. Comparison of ripple for torque and current for stator total harmonic distortion attained using different PMSM DTC techniques 6. CONCLUSION
Table 3. Comparison of ripple for torque and current for stator total harmonic distortion attained using different PMSM DTC techniques 6. CONCLUSION
Figure 13. PMSM DTC using a neural network that accounts for current for stator and ripple for torque
Figure 13. PMSM DTC using a neural network that accounts for current for stator and ripple for torque
Related topics:
Fuzzy Logic ControllerNeural Network Controller
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved