Application of Deep Neural Networks to Distribution System State Estimation and Forecasting

Abstract

Classical neural networks such as feedforward multi-layer perceptron models (MLPs) are well established as universal approximators and as such, show promise in applications such as static state estimation in power transmission systems. The dynamic nature of distributed generation (i.e. solar and wind), vehicle to grid technology (V2G) and false data injection attacks (FDIAs), may pose significant challenges to the application of classical MLPs to state estimation (SE) and state forecasting (SF) in power distribution systems. This paper investigates the application of conventional neural networks (MLPs) and deep learning based models such as convolutional neural networks (CNNs) and long-short term networks (LSTMs) to mitigate the aforementioned challenges in power distribution systems. The ability of MLPs to perform regression to perform power system state estimation will be investigated. MLPs are considered based upon their promise to learn complex functional mapping between datasets with many features. CNNs and LSTMs are considered based upon their promise to perform time-series forecasting by learning the correlation of the dataset being predicted. The performance of MLPS, CNNs, and LSTMs to perform state estimation and state forecasting will be presented in terms of average root-mean square error (RMSE) and training execution time. An IEEE standard 34-bus test system is used to illustrate the proposed conventional neural network and deep learning methods and their effectiveness to perform power system state estimation and power system state forecasting.