Predictive models aid energy efficient management
Energy efficient management of building systems will play a major role in minimising energy consumption and costs in the future, since a large part of today’s energy is consumed in buildings. The use of model predictive control (MPC) together with weather and occupancy predictions is an effective approach for achieving significant energy savings. Control of power grids is also of major concern due to the increasing number of renewable energy sources, which can cause additional variability in the power flow. This challenge can be mitigated by making use of additional storage by exploiting the thermal energy stored within the buildings itself. The EU-funded SMPCBCSG (Stochastic model predictive control, energy efficient building control, smart grid) project therefore investigated predictive computer models that can be applied to buildings and electricity grids. The Research Fellow developed, implemented and tested a new energy efficient control strategy for buildings on a test-bed at UC Berkeley, USA. They also formulated tractable dynamic power flow problems involving a stochastic formulation that included buildings as additional storage in the formulation. Challenges presented by efficient energy management in buildings and power grids were addressed using new stochastic formulations of MPCs based on martingale probability theory. The initial constraint formulation was based on the so-called scenario approach, which was then adapted online to constraint tightening based on the empirical constraint violation probability. A stochastic control strategy for buildings was developed that can also be used for provision of frequency regulation for the power grid. It included chance constraints due to uncertainty of the weather as well as robust constraints to ensure reserve provision for all possible frequency deviations. The new stochastic MPC method was tested for application to power grids. The results showed that the violation probability converges to the desired level, thus cost savings can be achieved. Furthermore, different analytical chance constraint reformations were compared in a security constrained optimal power flow formulation. SMPCBCSG investigated advanced and novel control solutions to manage uncertainty, large-scale systems, distributed systems, and predictions of the future system behaviour. Recent advances in random convex programming were coupled to MPC to develop a novel stochastic MPC formulation that could be applied to large-scale systems involved in building control and power grids.
Keywords
Energy efficient management, building systems, model predictive control, power flow, stochastic model, SMPCBCSG, smart grid, martingale probability theory
