Ambient Intelligence (AmI) Assisted Passive Ventilation in Mixed-Use Micro Apartment During SARS-CoV-2 Pandemic

Abstract

During recent and ongoing pandemic circumstances, a lot of architectural spaces were adapted for use not designed for. Besides ergonomics and comfortable furniture, occupational health hazards include more indoor air pollution induced by functionally over-saturated architectural spaces. This paper discusses options and proposes an algorithm to improve air quality inside mixed-use micro apartment using low energy consumption embedded artificial intelligence (AI) systems to assist users in passive ventilation usage. Through data collected for observed case, algorithm is explained and tested both in terms of feasibility in low power embedding and energy efficiency annual savings by using assisted passive ventilation. Air pollution and progressively unsustainable old, built-in materials and infrastructural systems in existing buildings with limited to none energy upgrade options need solutions for maintaining comfortable and healthy indoor environmental conditions. Proposed low power embedded, ambient intelligence system provides solutions for such architectural spaces. Case study included a variety of parameters in a complex physical model, and through data feature engineering most influential parameters were chosen. Time series forecasting for predictive maintenance of air quality and built-in materials was tested through three different models: ARIMA, Facebook’s Prophet and Tensorflow recurrent neural network (RNN) with gated recurrent units (GRUs). Machine learning algorithm (TinyML) was deployed to Arduino Nano 33 BLE Sense microcontroller board in testing phase, to prove simplicity and feasibility of chosen AI neural network. Validation is provided through simulation on collected data, to show ventilation energy savings by using AI assisted passive ventilation.