Regional wind power prediction method and device and medium

The invention belongs to the technical field of wind power prediction, discloses a regional wind power prediction method and device and a medium, and provides a coefficient index for judging time relevance and volatility strength, performs characteristic analysis on a wind power plant, classifies the wind power plant into four conditions, and designs different prediction methods. For a wind power plant with strong time relevance and strong volatility, data dimension reduction is carried out by adopting empirical mode decomposition noise reduction and principal component analysis, and prediction is carried out in combination with a CNN-LSTM model. For wind power plants with weak time-strong waves, strong time-weak waves and weak time-weak waves, BiLSTM, a random forest and a liquid neural network are respectively adopted, and the prediction precision is optimized through abnormal value processing, data filling and feature selection. And finally, performing prediction by selecting bagging ensemble learning, ada