Previous (Research and policy development)   |   Start   |   Next (Education and career)

Why do BPS?

Digital twin for operating control and virtual sensing

The term ‘digital twin’ has been used in an almost inflationary way in connection with buildings since the beginning of the 2020s. But if almost any digital model can be called a digital twin, the term will quickly lose its attraction. Hence, we would like to introduce a levelized categorization scheme for building performance digital twins:

Level 0: Online access to and visualization of sensor data in the context of a 3D geometrical model, cannot really be called a Digital Twin. There is no behavioral model of the building at all that could help to interpret the measured data.

Level 0: 3D Model + IOT signals

Level 1: The term ‘digital twin’ is often used in connection with a new operational monitoring system that integrates AI engine into the loop, that can be trained to provide system control. It enables some use for demand side management (load shedding and shifting), but as there is no physical understanding of the building, off-design behavior is generally unavailable.

Level 1: 3D Model + AI Engine

Level 2: This is the first level, where an online 3D white box simulation model enters the stage. Some of the sensed signals (e.g. weather data, BMS control signals, and electricity use) are used as input in open loop for driving the BPS. This is the lowest level, where not only better fault detection and diagnostics, but also a meaningful comparison of measured vs. simulated signals can be achieved. A BPS model can be calibrated by searching for the model parameterization for which the best match can be found. This in turn enables model-based normalization of energy use, i.e. to reliably separate between reference and actual building operation. Building occupancy patterns, weather or control settings can deviate from what was planned, but in the calibrated BPS model, these inputs can be “normalized”.

Level 2: 3D White Box Simulation

Level 3:  A correctly tuned model, that is made to follow the building in real-time based on a few measurements - often key temperatures, energy supplies and system flows – can provide virtual sensing. All model variables can be measured indirectly, also those that are in fact hard or even impossible to measure, such as PPD (Percentage of people dissatisfied) or other indoor climate indicators. Off-line what-if analysis based on a continuously calibrated model provides additional value. This way, changes in the systems (or settings) can be fairly evaluated also in terms of their actual impact on building utility and consequent risk for tenant dissatisfaction. True optimization of cost vs. benefit can take place.

Level 3: Building Tracker

Level 4: Not only a limited number of sensed signals, but also the entire BPS model as an accurate replica of the real building, can be used for operating model predictive control. Predicted inputs for both weather, occupancy, and loads are used to compute optimal control parameters for the building energy management system for (usually) the next 24h.

Level 4: Model Predictive Control

Previous (Research and policy development)   |   Start   |   Next (Education and career)