A recursive modelling technique applied to the model predictive control of fluid filled heat emitters

Based on the recent emergence of Controlled Radiator Valve (CRV) components, the paper considers the research, development, application and benefits of a modern control methodology to improve the heating efficiency of domestic dwellings. In particular, the problem of efficient temperature control, is formulated as a model predictive control scheme employing a parameter matching technique. A key contribution of the paper is the development of an on-line modelling method, which, in contrast to previously reported techniques, requires no prerequisite knowledge of the thermodynamic behaviour of a given controlled zone and a training period of only 48 h. Moreover, it is shown that excellent performance is obtained without the normal requirements for measurements of site weather or input from other external sources of weather data, thereby reducing system cost and complexity. The proposed techniques are applied in a controlled zone using a BS EN 442 oil filled heat emitter, whose input power is closely controlled using a PWM power converter within an instrumented test cell, and also in an occupied dwelling. Results demonstrated MPC can be implemented in a dwelling with minimal perquisite modelling and still achieve set point tracking when compared to more conventional solutions resulting in an energy saving of up to 22%.