Test Bench for Evaluation of Machine Learning Algorithms Applied to PQ Parameters Classification

In this paper, the test bench for evaluation of the machine learning algorithms, designed for an application in smart grids, will be presented. Further, the prototype is developed and machine learning algorithms for the classification of the voltage sags, based on a Long Short-Term Memory (LSTM) neural network are applied. The computation (processing) speed of algorithms is measured, and the proper classification is verified with synthetic signals. Also, network communication of the prototype device with the remote access device, via the web is tested, and response times and online delay are measured. The results of all tests and further plans for full laboratory setup development are shown.