A control method for time-delay system using improved genetic algorithm and adaptive smith predictor

Considering unknown systems with time-delay, a neural network approach for on-line parameter estimation is presented The unknown steady-state gain and time-delay of the systems are estimated on-line by neural network, and they are used to modify parameters of Smith predictor in real-time. An improved genetic algorithm is proposed in this paper This algorithm is an on-line technique, and is used to automatic tuning the gain and the three learning rates of the neuron controller in real-time. This method can be applied for slow time-varying and uncertainty systems with time-delay. Simulation results show that the method is efficient and practical.