Assessing simplified and detailed models for predictive control of space heating in homes
Published:
A model of a real system is required for predictive control to determine the best control sequence when disturbance forecasts and future system status are considered over a defined time horizon. The selected model should strike a balance between its accuracy and simplicity. This paper presents a comparison between different modeling approaches for predictive control of space heating. The case study is electric baseboard heating in homes within cold climate regions with the objective of reducing peak electricity demand (and saving costs if tariffs include a peak power charge). Detailed TRNSYS models of the selected house are developed and predictive control is implemented by using GenOpt as the optimization tool. This approach is compared with optimal predictive control algorithms based on simpler models. These models are obtained by parameter identification using data generated from the detailed TRNSYS models. Both approaches use perfect forecasts for the occupancy and the weather data in order to focus the analysis on model differences. Results show that MPC can deliver a significant reduction in power demand during on-peak periods with both modelling approaches (55% with detailed model, 33% with simplified model). The detailed model delivers significantly better savings but implies a calculation time that is more than 2 orders of magnitude higher. The potential of both approaches is discussed in the context of residential heating control to support a smart grid.
Recommended citation: Zhang K, Roofigari N E, Quintana H, Kummert M (2014). Assessing simplified and detailed models for predictive control of space heating in homes. In: Proceedings of 9th International Conference on System Simulation in Buildings, Liege, December 10-12, 2014, pp. P04.1-21.