Abstract: To address the issue of dynamic changes in grid operation constraints caused by wind power fluctuations, a dynamic assessment system for wind power grid integration capacity based on wind power forecasting is proposed. An improved LSTM multi-scale forecasting model is developed, incorporating sensitivity correction and dynamic constraint feedback mechanisms, forming a closed-loop process of forecasting-optimization-verification. Validation using typical grid data from Shandong Zhihui Electric Power Company shows that the designed system improves the acceptance power by approximately 13% under multiple load scenarios. When the forecasting error is controlled within ±5%, the voltage stability index reaches 0.94, demonstrating strong real-time performance and engineering application value.