Probabilistic Forecasting for Dissolved Gas Concentrations in Transformer Oil Based on the Bayesian Temporal Convolutional Network

Accurate prediction results of the dissolved gas concentrations in transformer oil can provide important information to support running state evaluation. The traditional methods can only output the deterministic forecasting results, lacking the ability to capture the existing uncertainties. To address this issue, this paper proposes a probabilistic forecasting method based on the Bayesian temporal convolutional network (BTCN). The TCN is used as the basic model structure to extract hidden features of the input historical gas data. The Bayesian theory is introduced to guide the training process of the model parameters. The experiment results show the effectiveness of the proposed method.