Highlight
The data model for Apollo is very complex. One optimization may involve tens of thousands of different time series. The data model includes the asset models with time series attributes describing, for example, input data required for optimizations and optimization results.
The basic process for creating operative optimization requires creating a snapshot of the inputs, running the optimizations, and generating predefined reports. This process can be fully automated and is based on templates.
For operative analysis, there is a need to analyze different combinations of time series and results, different detail levels of the results, for different asset models and their combinations, and so on. For this purpose, a domain-specific visual templating language was created, which makes it theoretically possible to create an endless amount of different visualizations. In practice, there can be dozens of charts and tables that can be browsed on demand and used to export reports. A typical operative analysis can result in a few GB of data being produced.
Ultimately, Apollo is an expert system that requires a lot of flexibility for the end-user - i.e. all the bells and whistles potentially available – a demanding IT infrastructure and many software performance optimizations – even for a single user. In practice, due to automatically scheduled case creations, optimizations and report generations, the system is technically demanding even without any users at all.
Fortum Apollo presented a very complex and interesting set of technical challenges to solve. We look forward to more challenges of this magnitude in the future.