Data-driven Assessment of Power System Reliability in Presence of Renewable Energy

The penetration of renewable energy sources (RES) and energy storage systems (ESS) in the modern-day power grid is increasing at a fast pace. However, reliability assessment of power systems using traditional methods has become a challenging task due to the interdependencies between RES like wind and solar, ESS, and the load. This paper proposes a new method based on artificial neural networks (ANN), a data-driven technique, for reliability assessment of a power system by estimating the parameters of the ANN. The hourly generation data of the distributed and conventional generators are considered to be the features or the input variables. A recurrent neural network based classification algorithm is trained to determine system responses to changes in system conditions. The data required for training and testing the learning algorithm is generated using sequential Monte Carlo simulation. The IEEE Reliability Test System is utilized for testing and validating the proposed approach. The results indicate that the learning algorithm can model the temporal relevance between different system variables for successful reliability assessment of the system.