Astronomical observations are critically dependent on precise, localized sky condition forecasts, a need often unmet by conventional weather prediction methods. This paper introduces an IoT-based intelligent system for hyperlocal sky condition forecasting, specifically tailored for astronomical observatories. The system leverages a comprehensive sensor network, including a Sky Quality Meter, to continuously gather hyperlocal environmental data. This data serves as input for a multi-output Long Short-Term Memory deep learning model designed to concurrently predict the probability of precipitation and SQM values. Experimental results, based on data collected at the Quy Nhon Observatory, demonstrate the system’s high efficacy. The model achieved a precipitation forecast accuracy of 97.77% and an R-squared value of 0.981 for SQM prediction. This research provides a significant contribution by offering an automated and accurate forecasting tool that can substantially improve astronomical observation planning, optimize resource utilization and enhance the protection of sensitive astronomical equipment. The proposed system presents a cost-effective and robust solution for observatories requiring specialized, real-time sky condition assessments.