CX2 Algorithm

CX2 Algorithm is the active layer on top of CX2 Open. Where CX2 Open collects, structures and makes building data available, CX2 Algorithm is meant to use that data to continuously optimise the building's operation.

CX2 Algorithm is not meant to replace the building management system. Ordinary control stays local in the building's BMS: PID control, valve and damper control, frost protection, safety functions and the other functions the BMS already handles well. CX2 Algorithm is intended for the functions that are typically hard to get working satisfactorily in a traditional BMS, because they require large amounts of data, history, continuous adaptation or calculations across several systems.

One example is optimal start after night setback. The BMS can start the heating at 05:43 without any trouble. The hard part is calculating that 05:43 is the right time tomorrow. That requires knowledge of the building's current temperatures, historical heating curves, outdoor temperature, weather forecast, the building's thermal properties and the desired comfort level.

Another example is the supply temperature in a mixing circuit. The BMS continues to control precisely to the given setpoint. CX2 Algorithm calculates what the setpoint ought to be, based on the building's actual demand, current room temperatures, history and possibly weather forecasts.

A central principle is that CX2 Algorithm must be predictable and controllable. AI can be used to analyse building operation, find patterns and suggest improvements, but the continuous influence on operation happens through well-defined algorithms with known limits and functions.

AI helps understand the building. CX2 Algorithm optimises the operation. The BMS performs the control.

The algorithms are not developed for one particular BMS. CX2 describes the building in Brick Schema, so the algorithms work against a shared semantic model rather than vendor-specific point names. That makes it possible to move a good algorithm from one building to another. Portability does not mean the algorithm can be used uncritically, though: buildings respond differently, and each algorithm must either be configured for the specific building or learn the building's actual behaviour before it is allowed to influence operation. An optimal-start algorithm can be the same across many buildings, but its model of heating time must be tuned to each building.

Brick makes the algorithm portable. Data makes it building-specific.

In the longer term the ambition is for CX2 Algorithm to become a library of reusable operational algorithms for optimal start, adaptive setpoints, predictive operation, demand-based operating hours and more. It also opens the door to collaboration with universities and other research environments: new algorithms can be developed and tested in one place and then used across many buildings through CX2's shared semantic data layer.

So is the idea that AI should run the building? No, that is not the intention. AI opens a great many doors, but we are not ready to hand the controls to a language model.

Does not exist yet.