Long-term wind speed forecasting is an important research topic in many fields, including power market restructuring, energy management, and wind farm optimal design. However, precise values alone are not enough to fully describe the changing trend of long-term wind speed forecasts. In contrast, interval variation and semantics can provide more comprehensive long-term wind speed forecast information. Aiming at the drawbacks of numerical forecasting of wind speed changes, this paper proposes a hybrid forecasting model that combines fuzzy cognitive maps and granular computing for forecasting. Specifically, the granular computing is used to granulate the data, and then the fuzzy cognitive map is used to predict the processed data. Compared with BP neural network, simulation experiments show that the proposed algorithm shows better performance in predicting interval values and semantics. This study highlights the importance of incorporating interval variation and semantic representation into long-term wind speed forecasts, and the proposed model can serve as a decision-making tool for wind energy industry stakeholders.