Neural-network-controlled single-phase UPS inverters with improved transient response and adaptability to various loads

This paper proposes a neural-network control scheme for the inverters of uninterruptible power supplies (UPS) to improve their transient response and adaptability to various loads. Two simulation models are built to obtain example patterns for training the neural network. One is a multiple-feedback-loop controller for linear loads, and the other is an idealized load-current-feedback controller specially designed for nonlinear loads. The latter controller has a built-in reference load model, and the load current is forced to track this reference. Example patterns under various loading conditions are used in the off-line training of a selected neural network which is made as simple as possible to reduce the calculation time. When the training is completed, the neural network is used to control the UPS inverter on-line. The development of example patterns and training of the neural network are done using MATLAB and SIMULINK, and the verification of results is done using PSpice. It is found that the proposed neural-network-controlled inverter can provide good sinusoidal output voltage with low total harmonic distortion (THD) under various loading conditions, and good transient responses when the load changes.