Data-Driven Estimation of Time-Varying Stochastic Effects on Building Heat Consumption Related to Human Interactions

Within the field of statistical modelling and data-driven characterisation buildings’ energy performance, focus is typically on parameter estimation building envelope systems. Less has been put stochastic human effect consumption. We propose a new method for estimating thermal properties while, in parallel, time-varying effects caused by humans’ interactions with building. do that combining smooth, non-linear formulation signature known from literature hidden state formulated as random walk to describe The demonstrated data obtained autumn 2019 late spring 2021 900 m2 newly built school located south Oslo, Norway. demonstration case shown model accuracy increases bias decrease when cross-validated. estimated also resemble combined mechanical natural ventilation pattern controlled users operational staff. These have increased total heat loss expressed kilowatts per kelvin around 50% over course one year before COVID-19 pandemic after its outbreak.