Data-Driven Virtual Replication of Thermostatically Controlled Domestic Heating Systems

Thermostatic load control systems are widespread in many countries. Since they provide heat for domestic hot water and space heating on a massive scale the residential sector, assessment of their energy performance effect different strategies requires simplified modeling techniques demanding small number inputs low computational resources. Data-driven envisaged as one best options to meet these constraints. This paper presents novel methodology consisting combination an optimization algorithm, two auto-regressive models loop algorithm able virtually replicate thermostatically driven systems. combined strategy includes all controlled modes governed by set point temperature enables automatic consumption impact multiple scenarios. The required limited available historical readings from smart thermostats external climate data sources. has been trained validated with sets coming selection 11 thermostats, connected gas boilers, placed several households located north-eastern Spain. Important conclusions research that can estimate decay when is off well needed reach comfort conditions. results also show estimated median savings 18.1% 36.5% be achieved if usual schedule lowered 1 °C 2 °C, respectively.