Power Quality Disturbance Detection using Autoencoder Neural Network

Power Quality is a vast and critical field of study in the power system dealing with the conformance of the sinusoidal current and voltage waveforms with the healthy nominal system requirements as governed by local and international standards. Power Quality Disturbances (PQD) are aberrations in the voltage and current waveforms under question and thus their existence has a pernicious effect on all the loads feeding on the grid. Thus, monitoring and measuring devices in large scale power systems produce large amounts of time series data which facilitates the need to develop algorithms that can quickly analyze and detect power quality disturbances to apply and develop the appropriate measures to avoid long term detrimental outcomes to consumers and added costs onto the grid operators. Therefore, this paper proposes an anomaly detection approach using an autoencoder neural network tested on odd harmonics, voltage sag and voltage swell as well as their combination in a sample power system signal.