Probabilistic Prediction of PV Ramp Events Based on LSTM-Bootstrap

In recent years, renewable energy has received widespread attention. However, due to the impact of climate change, the output power of photovoltaic (PV) power generation is highly variable, which poses a challenge to grid scheduling. The PV power ram event, i.e. the drastic change of PV power in a short period of time, has a particularly obvious impact on the power grid. This paper first summarises the current state of PV ramp-up research. Then, three common definitions of PV ramp-up events and the definition chosen in this study are described in detail. Then, a new ramp prediction method is proposed: first, the mRMR feature selection method is used to screen the input features; then, the LSTM network is constructed to predict the PV power; finally, the prediction intervals with different confidence levels are given in combination with the bootstrap method, and the probability of the ramp is determined accordingly. The effectiveness of the method is verified by simulation. This study provides new ideas for further improving the accuracy of PV ramp prediction. The experimental results show that by selecting the appropriate ramp threshold and determining the ramp probability, this method has achieved better results in terms of accuracy, hit rate, critical success index and other evaluation indicators.