Abstract:
Air pollution in the world increase due to the emission of pollutant gas from vehicles. To solve
this problem electric vehicle with traction motor instead of ICE is the emerging technology
but speed control of traction motor is the big challenge in electric vehicle. This thesis work
presents the design of a neuro-fuzzy sliding mode controller for switched reluctance motor
speed control applied to electric vehicle traction. The proposed Neuro-Fuzzy sliding mode
controller was trained with input and output data of the sliding surface, change in the sliding
surface, equivalent part inputs, and signum function output to replace the designed integral slid-ing mode controller signum function by ANFIS and its equivalent part by neural network. The
simulated integral sliding mode controller provided all of the data used to train the proposed
controller. To replace the signum function of the designed integral sliding mode controller with
ANFIS, the ANFIS was trained from the data of the sliding surface, change in sliding surface
and signum function output using a hybrid learning algorithm, and to replace the equivalent
part of the designed ISMC by a neural network, the neural network was trained from the data
of equivalent part inputs (reference speed, torque load, actual speed and speed error) and the
sliding surface using the Levenberg-Marquardt backpropagation algorithm. The stability of the
designed integral sliding mode controller is proven using the Lyapunov stability criteria. The
performance of the proposed controller was compared with ISMC and ANFIS sliding mode
controllers using MATLAB\Simulink simulation results with different operating conditions.
It is observed from the simulation results of the system using ISMC, ANFIS-SMC, and N-ANFIS-SMC controllers that with 2000 rpm step reference speed, the percentage overshoots
were 0.725%, 0.725% and 0.54% respectively and the N-ANFIS-SMC controller reduces the
torque ripple by 34.1% and 43.2% than using ANFIS-SMC and ISMC controllers respectively.
This shows that the N-ANFIS-SMC has decreased overshoot and torque ripple. Also, the chat-tering effect is minimized and the speed response reaches its setpoint value quickly in the
N-ANFIS-SMC controlled traction motor. This shows the effectiveness of the designed N-ANFIS-SMC controller. On the other hand, ANFIS-SMC shows better performance than ISMC
for vehicle traction motor. This is because the discontinuous part of ISMC is replaced by AN-FIS to make it continuous.
Key Words: ANFIS, battery, electric vehicle, neural network, sliding mode control, switched
reluctance motor.