In recent years, offshore wind farms, a crucial component of renewable energy, have attracted widespread interest and development worldwide. Nevertheless, offshore wind farms face a variety of meteorological risks during operation, including wind speed fluctuations, strong winds/storms, and extreme storms, which can have a significant impact on the safe operation and stable power generation of wind turbines. Existing methods for predicting meteorological risk frequently lack dynamism and adaptability, failing to meet the requirements of practical applications. This paper proposes a dynamic meteorological risk prediction method for offshore wind power based on hidden Markov models to address this issue. First, we propose four risk states based on the operation of offshore wind turbines under different wind speeds and meteorological conditions: extreme storm events, extreme ocean events, wind speed fluctuations, temperature fluctuations. Then, we construct state transition matrices and output matrices by collecting actual observational data (such as wind speed and wind direction) and combining expert experience and historical events. Finally, we use hidden Markov models to predict the risk states of offshore wind turbines in a dynamic manner. This paper uses artificially generated data to test and compare the performance of the proposed method, demonstrating that it significantly outperforms traditional Markov models and naive Bayes models in state prediction accuracy and is adaptable to some degree. In practical applications, the method can be continuously adjusted and optimized to improve prediction accuracy. By applying the dynamic meteorological risk prediction method for offshore wind power to actual scenarios, wind farm operators can receive real-time information about risks and take the necessary precautions to ensure the safe operation and stable power generation of wind turbines.