Optimal supervisory control of renewable energy systems in buildings

The project (GR/S30467) led by Professor Vic I Hanby has been sponsored by The Engineering and Physical Sciences Research Council (EPSRC).



Model-based control is emerging as an important research area in building control systems, which are having to cope with the control of an increasing variety of systems. With this form of control the system runs, in real time, a mathematical model of the system and uses the output of the model to inform control decisions. Where the characteristics of the building and its systems are dynamic in character, model-based control becomes a challenging proposition. This research project was targetted on the use of renewable energy systems in buildings which also incorporate thermal storage, both in the fabric of the building and also in an active thermal store. As the availability of energy and the state of the building/system vary with time, conventional static design procedures are unlikely to be able to take full advantage of the potential contribution of renewable sources.

The research was based on a real building, the Brockshill Environment Centre in Oadby, Leicestershire. The building was completed in 2001 and was constructed to showcase renewable energy and other aspects of sustainable practice. The systems at Brockshill include a wind turbine, ventilated solar photovoltaic panels, air- and water-heating solar collectors, an active 1000-litre thermal store and a dual fuel (oil and biomass) boiler. The control system devised at the design stage was based on rules derived from standard industry practice - this served as the benchmark for this study but it is important to remember this is an arbitrary datum and that there is not a reservoir of experience of devising control systems for these installations, especially in the UK.

In this project we developed a mathematical model of the building and its systems and calibrated this carefully with the data (temperatures, flow rates, control positions etc.) obtained from the installed building management system. The next stage was to combine this model with advanced evolutionary optimization procedures to determine a set of optimal control settings. This was done on a daily basis: an overnight computer run was made to provide a set of control signals for the following day: this schedule was then updated at fifteen-minute intervals during the day.

Initially, actual measured weather data was used as input to the model but in the final stages of the research, we incorporated weather prediction into the process, as would be necessary in a real-time situation. It was found that the optimal control schedules devised could perform significantly better than the benchmark, in terms of minimizing the 'back-up' energy which the building used. This suggested that our approach would be capable of improving the operation of complex energy systems in buildings at potentially small capital cost (additional computation power is needed to run the combined model/optimization).

It was not possible within the constraints of this project, to actually implement the optimal control in the building. However the project was effective in demonstrating the potential of optimal control, highlighting the remaining problems to be solved before implementation (principally the speed of computation) and potentially innovative strategies which could inform current practice.