No abstract

Occupants often perform many types of behavior in buildings to adjust the indoor thermal environment. In these types, opening/closing the windows, often regarded as window-opening behavior, is more commonly observed because of its convenience. It not only improves indoor air quality to satisfy occupants' requirement for indoor thermal comfort but also influences building energy consumption. To learn more about potential factors having effects on occupants' window-opening behavior, a field study was carried out in an office building within a university in Beijing. Window state (open/closed) for a total of 5 windows in 5 offices on the second floor in 285 days (9.5 months) were recorded daily. Potential factors, categorized as environmental and nonenvironmental ones, were subsequently identified with their impact on windowopening behavior through logistic regression and Pearson correlation approaches. The analytical results show that occupants' window-opening behavior is more strongly correlated to environmental factors, such as indoor and outdoor air temperatures, wind speed, relative humidity, outdoor PM2.5 concentrations, solar radiation, sunshine hours, in which air temperatures dominate the influence. While the non-environmental factors, i.e. seasonal change, time of day and personal preference, also affects the patterns of window-opening probability. This paper provides solid field data on occupant window opening behavior in China, with high resolutions and demonstrates the way in analyzing and predicting the probability of window-opening behavior. Its discussion into the potential impact factors shall be useful for further investigation of the relationship 3 between building energy consumption and window-opening behavior.

No abstract

No abstract

Recently, there is an emerging trend to apply deep reinforcement learning to solve the vehicle routing problem (VRP), where a learnt policy governs the selection of next node for visiting. However, existing methods could not handle well the pairing and precedence relationships in the pickup and delivery problem (PDP), which is a representative variant of VRP. To address this challenging issue, we leverage a novel neural network integrated with a heterogeneous attention mechanism to empower the policy in deep reinforcement learning to automatically select the nodes. In particular, the heterogeneous attention mechanism specifically prescribes attentions for each role of the nodes while taking into account the precedence constraint, i.e., the pickup node must precede the pairing delivery node. Further integrated with a masking scheme, the learnt policy is expected to find higher-quality solutions for solving PDP. Extensive experimental results show that our method outperforms the state-of-the-art heuristic and deep learning model, respectively, and generalizes well to different distributions and problem sizes.

Modeling of window behavior is a key component for building performance simulation, due to the significant impact of opening/closing windows on indoor environment and energy consumption. The predictions of existing models cannot well reflect actual window behavior, the prediction accuracy still needs to be improved. The Gauss distribution model is a new machine-learning technique which has achieved successful applications in many fields because of its special advantages (i.e. simple structure, strong operability and flexible nonparametric inference ability) compared to existing models. This paper presents results from a study using the Gauss distribution model to predict window behavior in office building. The data used in this study were from a real building located in Beijing, China, and covered two transitional seasons (from October 1 to November 15, 2014 and from March 15 to May 16, 2015), when natural ventilation was fully applied. When modeling, three types of input variables, i.e., indoor temperature, outdoor temperature and their combination were used. This work validates the importance of selecting suitable input variables when developing Gauss distribution model. This study also compared the prediction performance between the Gauss distribution modeling approach and the Logistic regression modeling approach, which is the most popular method used to model occupant window behavior in buildings. The results showed that Gauss distribution models could provide higher prediction accuracy, with 9.5% higher than Logistic regression model when using suitable inputs. This paper provided a novel modeling method that can be used to predict window states more accurately in office buildings.

No abstract

No abstract