Neural and Neuro-Fuzzy Controllers for UPS Inverter Applications

In this work neural and neuro-fuzzy controllers are developed for the inverters of Uninterruptible Power Supplies (UPS) to improve their transient response and adaptability to various loads. Idealized load-currentfeedback controller is built to obtain example patterns for training the networks. Example patterns under various loading conditions are used in the off-line training of the selected neural and neuro-fuzzy networks, which is made as simple as possible to reduce the calculation time. Each time the weights and biases of Neural Network (NN) parameters of adaptive node in Adaptive Neuro-Fuzzy Inference System (ANFIS) are updated using the back propagation algorithm to make the mean square error between the desired output and actual output less than the predefined value. When the training is completed, the neural and neuro-fuzzy controllers are used to control the UPS inverter online. Simulation is done for both linear model of UPS inverter and for thyristor model without controller and with PI, neural and neurofuzzy controllers. The simulation results shows that the developed neural and neuro-fuzzy controllers can provide good sinusoidal output voltage with low Total Harmonic Distortion (THD) under various loading conditions, and exhibits good transient performance when the load changes compared to without controller and with PI controller. Index Terms — UPS, ANFIS, NN, THD