**5. Conclusions**

Dispatching electricity consumption of SETS is an effective method for peak shaving in the power system. The prediction of thermal storage is a prerequisite for dynamic optimal dispatching, and the thermal storage can be used to shift times of resources. The CPS approach is proposed to predict power load, which promotes the application of the CPS in smart grids and ubiquitous power internet of things. This paper proposes a cyber–physical approach to predict the electricity consumption of SETS with consideration of the ambient temperature and electric behavior of the SETS into the PM. The CM adopts the BP neural network to calibrate the errors obtained in the PM. The 1MW SETS is established to validate the proposed cyber–physical approach. The simulation results show that when the CPM is compared with the PM, the MRE is reduced by 25.4%, and when compared with the CM is reduced by 4.8%. Using the CPM to calibrate the PM effectively improves the prediction accuracy. Conclusively, the prediction of SETS using the CPM is better than the individual PM and the CM alone.

The recommendations for future research are as follows.


**Author Contributions:** Conceptualization, H.J., J.Y. and H.W.; methodology, H.J., J.Y. and H.W.; software, H.J.; validation, H.J., J.Y. and H.W.; formal analysis, H.J.; investigation, H.J.; resources, H.J.; data curation, H.J. and K.T.; writing—original draft preparation, H.J.; writing—review and editing, H.J., H.W. and M.O.O.; visualization, H.J. and J.F.; supervision, J.Y. and H.W.; project administration, J.Y. and H.W.; funding acquisition, H.W.

**Funding:** This work was supported in part by the China Postdoctoral Science Foundation under Grant 2019M651144, in part by the Liaoning Provincial Department of Education Research Funding under Grant LQGD2019005.

**Conflicts of Interest:** The authors declare no conflict of interest.
