An artificial neural network model using outdoor environmental parameters and residential building characteristics for predicting the nighttime natural ventilation effect

The natural ventilation rate in bedrooms at night has been found to be insufficient at times. To create good indoor air quality with minimal energy consumption, we need to know when the ventilation rate will be insufficient. To achieve this goal, we monitored indoor CO2 concentrations in bedrooms in 24 apartments, together with related outdoor parameters. With this dataset, we built an artificial neural network (ANN) model to predict when the ventilation rate would be insufficient. The sample sizes for the training set and test set are 2760 sample days and 690 sample days, respectively. The overall accuracy levels of this ANN model are 80.2% and 79.3% for the training set and test set, respectively. According to the model, the indoor CO2 concentration level is significantly affected by the building floor on which the apartment is located and nighttime outdoor temperature. Apartments on an upper floor usually had lower probability of the indoor CO2 concentration reaching 1000 ppm than apartments on a lower floor. Furthermore, the probability dropped to its lowest point at 15–20∘C, and it rose significantly when the nighttime outdoor temperature increased or decreased. Increasing wind strength decreased the probability only when the outdoor temperature was greater than approximately 20∘C. In contrast, when the temperature was low, increasing wind strength may have caused the probability to increase because of the uncomfortable sensation created by increasing wind velocity. Throughout the year, the probability is estimated to be high in the middle of summer and winter.