The US Census Bureau reported the information that over ten thousand large commercial warehouses were being operated, and that over five hundred thousand workers were employed in related services in 2008. According to the US Bureau of Labor Statistics, the number of persons employed increased as nearly nine hundred thousand in November 2016, which implies the fact that demands of warehouses will increase in keeping with the growth of logistics industry. However, the necessity for energy savings was recognized as less important because the average energy use intensity of warehouses in the US was 55% and 70% less than those of office buildings and retail stores, respectively. Also, the improvement of indoor thermal environment for workers was often overlooked in comparison with safety, speed, and space efficiency. This research proposes a study for a warehouse building to mitigate both thermal dissatisfaction and energy use through the network based real-time analysis. In order to optimize heating and cooling supply, an algorithm for simultaneous control of the amount of air and its temperature is designed, and a neural network model that learns the algorithm is generated. Also, an inner algorithm for thermal comfort analyses real-time temperature levels and rectifies the model's control signals to mitigate thermal dissatisfaction. By comparing results, this research concludes advantages of a neural network model with estimating thermal comfort. The model reduces thermal dissatisfaction by 21.2% and saves energy use by 6.4% in comparison with the conventional thermostat on/off controller equipped in most buildings. Without compromising thermal comfort for workers, the proposed model that consists of two independent structures for optimizing supply air and estimating thermal comfort can contribute to the improvement of thermal performance for warehouses. (C) 2018 Elsevier B.V. All rights reserved.