Several studies have indicated a significant potential in using Model Predictive Control (MPC) of space heating for demand response purposes. The performance of the MPC depends on the predictive performance of the embedded control model. The studies often employ black- or grey-box control models; however, no previous studies consider whether a black- or grey-box model is more robust against weather changes. To assess this, the simulation-based study reported in this paper analysed how the predictive performance of black- and grey-box models trained with different input–output datasets from a certain period of a year is affected when subject to weather conditions in other periods of the year. The predictive performance of the grey-box models was slightly better compared to the black-box model. Furthermore, the grey-box models were slightly more robust to changes in weather data. Future studies should investigate whether the differences have practical significance in relation to MPC.

几项研究表明，将空间供暖的模型预测控制 (MPC) 用于需求响应目的具有巨大的潜力。MPC 的性能取决于嵌入式控制模型的预测性能。这些研究通常采用黑盒或灰盒控制模型。然而，以前没有研究考虑过黑盒模型还是灰盒模型对天气变化更稳健。为了评估这一点，本文报告的基于模拟的研究分析了使用一年中某个时期的不同输入-输出数据集训练的黑盒和灰盒模型的预测性能在受到其他时期的天气条件影响时如何受到影响年。与黑盒模型相比，灰盒模型的预测性能略好。此外，灰盒模型对天气数据的变化更加稳健。未来的研究应该调查这些差异是否对 MPC 具有实际意义。