Training a new model takes something like 30 seconds. Sometimes a bit quicker, sometimes a bit slower, depending on the size of your project and how busy the cloud service is. But it's around that order of magnitude. Is recreating the model really unacceptable? Do you really need the model to be idle and available 24/7 for a year?
If the answer is "yes", and you absolutely need the model to be present all of the time, then I think your main option would be to use your own IBM Cloud API key.
At the moment, I assume you're using mine, sharing it with many other schools and coding groups who also use it for their projects. Deleting your model frees up space for them to create one for their students. Leaving your model around persistently would be one fewer student in another school or code club who can create a model for their project. If you use your own API key, you don't need to share it with other groups.
Kind regards
D