The strong wind weather phenomenon due to tropical cyclone leads to abnormal output power variation of offshore wind farms

forms into a wind power ramp phenomenon and seriously threatens safe operation of the main power grid on shore. Based on this

a forecasting method for offshore wind power ramp events based on key creep causes is proposed by analyzing and constructing creep causes considering typical meteorological factors of tropical cyclone. Firstly

the influence of typical meteorological factors on wind power and its ramp events under tropical cyclone conditions is analyzed

and a wind power ramp events attribution model and key features of ramp events causes are constructed. Then

based on the key characteristics of s ramp events causes

a cost-sensitive logistic regression model

a weighted plain Bayesian model and a cost-sensitive random forest model are adopted to construct a base model for direct ramp events prediction. Finally

the integrated learning method based on bagging is used to improve the robustness and prediction accuracy of the model. The numerical results show that

compared to the feature triplet and the feature quintuples

the accuracy of the base model can be improved by 16.63%

10.94%

and 2.52%

respectively. Moreover

compared to the base model

the combined model can guarantee the robustness under the signal-to-noise ratio from 5 dB to 25 dB

and the accuracy of the combined model reaches up to 93%at the signal-to-noise ratio of 25 dB

which confirms the applicability of the proposed model in unbalanced data sets and noisy environments.