IEEE-519-based real-time and optimal control of active filters under nonsinusoidal line voltages using neural networks

In this paper, a fast and simple neural network (NN)-based control system for shunt active power filters operating under distorted voltage conditions is developed. The proposed system is an enhanced version of the optimal and flexible control (OFC) strategy with very fast and simple structure. In the proposed system, the time-consuming and complex nonlinear optimization algorithm required by OFC is replaced by a simple 3-layer perceptron NN. The NN is trained off-line using some random data based on the IEEE-519 Standard, while it can be used for a very wide range of new voltage waveforms in practice. The proposed system has been developed after introducing a new version of the OFC strategy in a-b-c frame of reference. This system satisfies both theoretical and practical requirements. Several simulation results using MATLAB toolboxes under highly distorted and unbalanced voltages have been provided to validate the ability of the proposed control system.