Design and application of fractional order PID controller with adjustable coefficients by using adaptive and neuro-fuzzy systems

The use of fractional order controllers (PIλDμ) instead of classical integer order controllers is becoming increasingly common and preferred. By using fractional order systems, system responses with high performance and robustness can be obtained. For fractional order controller design, control coefficient and parameters should be calculated in the most appropriate way. By using adaptive, fuzzy logic and artificial neural network control methods either together or separately, less design time, speed and robustness gained to the design of fractional order PID controllers with adjustable coefficients. In this thesis, firstly a fractional order controller design with adaptive model reference method was performed. After this stage the controller was designed with neural-fuzzy system structure. A fractional PID controller structure was constructed using the adaptive neural fuzzy inference system (ANFIS) for fractional order neural-fuzzy controller design. In this way, the advantages of fuzzy logic and artificial neural networks are used together, because these methods have the ability to self-learning, self-organizing and self-adjusting to the optimum point, online or offline. Artificial neural networks and previously trained ANFIS blocks are used as coefficients of the fractional PID controller designed in this thesis and the simulations of the designed controllers has been carried out. At the last stage a fractional order controller was designed with a auto-tuning neuron that uses the modified hyperbolic tangent function and the unit step responses were taken. At the end of the study, the computer simulations show that the controllers whose coefficients adjusted on-line have better performance, stability and robustness than the other controllers. Furthermore, the effort and time lost in the process of finding and adjusting the controller coefficients has been reduced to minimum level.