Typical Scenario Selection for Energy Storage Planning from Annual-Horizon Operation of A Large- Scale Power System with Renewable Integration

With long-duration energy storage technologies evolve, the cycling period of energy storage is no longer limited to a single day. The cross-day operation of energy storage engages in the regulation of renewable energy on longer time scales. Traditional methods for extracting typical days based on renewable energy output or the net load of the entire network have not adequately accounted for the operational modes and the flexibility of energy storages and units. In this study, a method for identifying typical days is employed, which utilizes optimal operational outcomes as the data foundation, thereby taking into account the operational characteristics of renewable energy, energy storage, and conventional units within the system. A machine learning approach combining supervised and unsupervised techniques is applied to extract a set of typical days that capture system operational features. This methodology has been validated within a 38-node network with high renewable penetration.