From model to control loop.
Between a plausible prediction and a change to cooling lie practical integration decisions. They are part of the engineering.
Observation establishes a baseline.
The starting point is the data the system actually provides. Measurement locations, timestamps, units and update rates need to align. Missing or stale values should be identifiable before they enter model evaluation.
Model predictions can then be compared with observed behavior. Steady operation alone is not enough: load changes and different environmental conditions reveal where a model fits well and where uncertainty increases.
Recommendation and actuation are separate steps.
A calculated control input can first be evaluated as a recommendation. Direct actuation additionally requires defined responsibilities, permitted ranges and suitable interfaces. The appropriate stage depends on the system and its operations team.
Limits must be part of the design.
Integration planning also covers missing data, a model operating outside its evaluated range or a lost connection to the system. These cases require behavior agreed with the operator and a clear path back to the previous operating mode.
The system diagram on this website describes this operating principle. Specific control inputs, deployment locations, approvals and fallback mechanisms are defined for each pilot project.