A direct prediction method for wind power ramp events considering the class imbalanced problem

Predicting wind power ramp events directly based on the historical event time series has drawn increasing attention recently. But class imbalance problem of significantly affects prediction accuracy events. In present study, a layer oversampling (LOS) method is proposed considering relation characteristics amplitudes and occurrence frequency Meanwhile, hybrid sampling error bootstrap-LOS (EB-LOS) by combining LOS with EB method. After balancing samples nonramp using different methods, backpropagation neural network (BPNN), long short-term memory (LSTM) methods are employed to predict data collected from eight farms. Comparison results proved that EB-LOS achieves best performance an average recall 0.8196 when BPNN model The also LSTM