Planning astronomy observations for telescopes is a very hard problem as it must deal with fully automatic operation, and dynamic rescheduling of observations based on changes in weather conditions, source visibility, technical failures, etc. Unlike state-of-the-art scheduling methods, planning observations requires intensive work to adapt the observation plans to the changing conditions of observation projects. Furthermore, some schedulers that use machine learning techniques require complex sample data of observation sequences, which are usually not available for most of the astronomy telescopes. In addition, since traditional scheduling methods are unable to self-organize, they usually require effort to optimize system parameters. In order to address these issues, in this work a new method is proposed to schedule astronomy observations projects. The approach uses artificial immune systems techniques in order to optimize observation plans and available resources according to real-time scientific priorities. Experiments using real and synthetic data on observation proposals and weather information, show the promise of the method when compared with traditional scheduling algorithms. The Atacama Large Millimeter/submillimeter Array (ALMA) is the biggest radio-interferometer telescope constructed in the Chilean Atacama desert. The scheduling system for ALMA considers a full automatic operation, and a dynamic re-scheduling of observations according to changing factors, like atmospheric conditions, source visibility, technical failures or targets of opportunity. This article proposes a new scheduling algorithm for ALMA based on immune system. It is verified against real data and focused in define a metric based on quality of the scientific output and instrument usage in a real world problem.