Abstract: Accurate ultra-short-term wind power forecasting is vital for optimizing grid scheduling and reducing operational costs. This paper addresses the continuous degradation of wind turbines during operation and its impact on wind power prediction. Initially, a unit status assessment is conducted using the fuzzy comprehensive evaluation method. Subsequently, based on these assessment results, a multidimensional wind power prediction model is developed. By integrating the unit status evaluation with numerical weather forecast data, ultra-short-term wind power predictions are made using convolutional neural networks and long short-term memory networks. The results demonstrate that the proposed method effectively mitigates the interference of unit status on power prediction, thereby enhancing prediction accuracy.