Wind and PV Power Ramp Events Prediction Based on Long Short-Term Memory Network

With the increasing penetration rate of renewable generation, the challenges arising from its randomness and intermittent nature are becoming more pronounced. In particular, wind and photovoltaic power ramp events pose a serious threat to the secure and stable operation of power systems, and even lead to frequency instability and load shedding. Based on the long short-term memory neural networks (LSTM), this paper proposes a novel renewable power ramp events (RPREs) forecasting approach. First, filter the key tie-lines. Second, the LSTM network was driven by the known time series of wind, PV, load, and tie-line power to forecast the tie-line power for the next hour, thereby determining the adjustable capacity of the tie-line. Finally, the regulatory capability of the grid is considered to judge the occurrence of a power ramping event. Case study demonstrates that the proposed method can effectively predict power ramping events and meets the requirements of online assessment.