Predicting indoor air temperature and thermal comfort in occupational settings using weather forecasts, indoor sensors, and artificial neural networks

Climate projections of human thermal comfort for indoor workplaces

“…The approach is based on a combination of indoor sensors, ANN models, and climate projection data. Human thermal comfort and T i measured by low-cost sensors at 90 different workplaces were used as training data sets for ANN models predicting indoor conditions as a function of current and past outdoor weather based on the approach of Sulzer et al ( 2023 ). The workplace-specific climate projections were modeled based on simulations of 3-hourly conditions for the future time period 2070–2099.…”

“…The ANNs used the optimizer “adam” (Kingma and Ba 2017 ) with a default learning rate of 0.0001. The structure of the ANNs is the same as for the workplace-specific short-term indoor heat health warnings presented in Sulzer et al ( 2023 ) where further details on the technical aspects of the ANNs and their evaluation can be found, including the testing of different input variables and a variable importance analysis.…”

“…The target data used for the training of the ANN models were measured by a custom-made low-cost sensor system, the “Mobile Biometeorological System” (MoBiMet), from Spring 2021 to Autumn 2022 in the Upper Rhine Valley region: at 51 workplaces in the area of Freiburg, Germany; 25 workplaces in the area of Basel, Switzerland; and 14 workplaces in the area of Strasbourg, France (Sulzer et al 2022 , 2023 ). The majority of the workplaces used for this study are offices (53), but also workplaces in laboratories and workshops (8), production (7), storage and logistics (10), as well as agriculture and forestry (12) were equipped with MoBiMets.…”

“…In a previous study, we showed that low-cost indoor sensors can be used for nowcasting thermal comfort at different indoor workplaces (Sulzer et al 2022 ) and as training data for ANN models to forecast short-term indoor thermal comfort and issue indoor heat warnings (Sulzer et al 2023 ). The goal of this article is to expand the short-term forecasting approach of Sulzer et al ( 2023 ) to calculate workplace-specific climate projections using low-cost sensors and ANN models.…”

“…In a previous study, we showed that low-cost indoor sensors can be used for nowcasting thermal comfort at different indoor workplaces (Sulzer et al 2022 ) and as training data for ANN models to forecast short-term indoor thermal comfort and issue indoor heat warnings (Sulzer et al 2023 ). The goal of this article is to expand the short-term forecasting approach of Sulzer et al ( 2023 ) to calculate workplace-specific climate projections using low-cost sensors and ANN models. We present climate projections for 90 indoor workplaces located in the Upper Rhine Valley in Central Europe, which are classified into offices, laboratories and workshops, production, storage and logistics, and agriculture and forestry.…”