Artificial neural network model of discharge lamps in the power quality context

This paper presents a methodology for modeling high intensity discharge lamps based on artificial neural networks. The methodology provides a model which is able to represent the device operating in the frequency of distribution systems, facing events related to power quality. With the aid of a data acquisition system to monitor the laboratory experiment, and using $$\text{ MATLAB }^{\textregistered }$$ software, data was obtained for the training of two neural networks. These neural networks, working together, were able to represent with high fidelity the behavior of a discharge lamp. The excellent performance obtained by these models allowed the simulation of a group of lamps in a distribution system with shorter simulation time when compared to mathematical models. This fact justified the application of this family of loads in electric power systems. The representation of the device facing power quality disturbances also proved to be a useful tool for more complex studies in distribution systems. © 2013 Brazilian Society for Automatics - SBA.