Technical Program

Buildings are responsible for around 40% of total final energy. Many initiatives have been developed to build more sustainable buildings: for instance, in European Union, all new building have to be nearly zero energy by 31 December 2020. 

In this context, in addition to material and energy resource researches, the building sector is interested in developing more efficient building management systems (BMS) that could adapt building controls in real time. Adapting controls to real-time disturbance (occupancy, weather, etc.) given the large amount of constraints (thermal comfort…, equipment capacities, etc.) and parameters (states of the building and HVAC systems) is a complex optimization problem but can lead to a significant reduction in energy consumption. 

Model Predictive Control (MPC) can provide a solution to this complex optimization problem. MPC is a class of algorithms that can exploit the historic measurements of a building to predict the future performance and find an optimal control strategy based on these predictions. 

With regards to this, we are developing MPCPy, an OpenSource Model Predictive Control (MPC) platform that facilitates the testing and implementation of MPC for building systems. The software package focuses on the use of data-driven simplified physical or statistical models to predict building performance and optimize control. Four main modules contain objects classes to import data, interact with real or emulated systems, estimate and validate data-driven models, and optimize control inputs. 

Implementing and using Haystack semantic with MPCPy could be very useful, cost-effective and time-gaining as it would permit to easily connect any new building without having to reconsider all the database and sensors. Therefore, this presentation will discuss how the Haystack semantic can enhance MPCPy’s application across buildings.