Publication

Abstract : Smart grid is prone to many power quality disturbances, which can cause considerable damage to consumer equipment. Conventional protection schemes can only provide a reactive response to the occurrence of these power quality disturbances. Power quality disturbances like oscillatory transients might have very high voltage values, which would destroy the insulation of transmission cables or the electronic circuits inside the consumer end devices. Therefore, it will be highly advantageous to forecast the occurrence of these power quality disturbances so that immediate measures can be taken to mitigate them. The forecast determines if a new power load will introduce a specific PQ problem to the power grid. Traditional methods struggle with the complexity of the system and lack of efficiency. Hence, newer methods were introduced for faster and accurate operation. This paper proposes the latest Encoder-Decoder model to forecast the power quality disturbances and a hybrid Convolutional Neural Network Long Short Term Memory model to classify the disturbances. After categorizing the disturbances, adequate mitigation measures can be taken to reduce their impact. Various trials are carried out to provide an optimum model with specific network parameters and topologies. Over 1S sets of unique and combined PQ events were used to assess the performance of the proposed system. The proposed architecture proved accurate for forecasting and classifying power quality disorders in the smart grid.