Abstract: Compared with extreme weather, the impact of weather factors on the supply and demand balance of the power system is often overlooked, but events such as no wind and no sunlight for many days can also cause problems in keeping the balance. This paper proposes a supply and demand imbalance risk fast assessment method for new power systems empowered by weather data and deep active learning, which is more efficient and accurate than traditional methods. First, a power system production simulation model on a daily scale is established considering the multiple links of source, grid, load, and storage to analyze the system supply and demand imbalance under abnormal weather conditions such as no wind and no sunlight. At the same time, because of the shortcomings of traditional reliability indices, a new daily-scale distributed index is proposed, and the risk curve is further adopted to describe the long-term risk of the system. Then, a power system risk fast assessment framework using deep active learning is proposed, a dual deep neural network coupled with a main risk prediction network and an error prediction sub-network is established, and the corresponding loss function and training process are constructed. Finally, the effect verification and comparison of various methods are conducted based on IEEE standard test cases, and the results verify its efficiency, accuracy, and scalability. This research proposes a novel and effective idea for fast risk assessment of new power systems.