Uninterruptible power supplies (UPSes) are used to provide back up power to critical loads, where even a temporary power failure causes large economic losses. The UPSes are used in applications like computer installations, process control in chemical plants, and general communication systems. The function of UPS is to provide a high quality sinusoidal voltage and for this it needs to be controlled. Conventional controllers cannot provide the exact sinusoidal output with the nonlinear loads. The artificial neural network controller having an advantage of adaptability to all loads. The neural network controller is built and trained by using patterns obtained from a simulated controller with multiple feedback loops having an idealized load-current reference. The weights and biases are downloaded to neural network controller and then it is used for online control of inverter. This is to achieve low total harmonic distortion and pure sinusoidal output.