A proactive energy-efficient optimal ventilation system using artificial intelligent techniques under outdoor air quality conditions

Highlights

• The deep learning model predicted future outdoor air quality in terms of PM₁₀.

• The proactive system was evaluated under several outdoor air quality conditions.

• The proactive system reduced energy by 8.68% while maintaining healthy indoor air.

• Annual greenhouse gas emissions could be diminished by 53,290 tons of CO₂.

• The proactive system could save the operating expenditure by $2,340,045 dollars.

Abstract

Passengers are directly or indirectly exposed to fine dust in indoor air of underground subway stations, which greatly affects the comfort and health of the passengers. However, conventional ventilation systems are manually operated, resulting in high energy consumption, and it is hard to consider the dynamic characteristics of indoor air quality due to the complex relationships between the environment of the subway station and climate change-driven outdoor air quality. Therefore, an energy-efficient ventilation optimization system based on deep learning and artificial intelligence (AI)-iterative dynamic programming was developed in this study for proactive environmental and economic maintenance of the underground ventilation system for a subway’s indoor air quality. The deep learning model predicted the next 24 h of the subway’s environmental status, and the AI-iterative dynamic programming searched a piecewise operational policy of ventilation flow rate for the same operational duration. Energy efficiency was improved by 8.68% while maintaining healthy indoor air quality for the passengers. The proactive optimal ventilation system for the platform of a target subway station presented a decrease of 96 tons of CO₂ per year to help address climate change and operating expenditure savings of up to $4217 dollars per year.

Introduction

The subway transportation system is an ideal sustainable transport solution to reduce greenhouse gas (GHG) emissions of urban mobility [1]; therefore, such systems have been rapidly applied to urban areas [2]. Although the initial capital costs for the system are relatively higher than those of other ground transportation systems, the subway transportation system has been selected as the preferred environmental system since it has high capacity, reliability, and environmental benefits compared to other ground transportation systems [3]. In Korea, nine subway lines with 320 stations have been installed in Seoul and are used by approximately 2.86 million people each day [4]. Therefore, high capacity and increasing usage of the subway system have increased attention on improving system efficiencies. Thus, it is urgent and significant to reduce the energy consumption of the subway system and to maintain its service quality with its social and environmental benefits [5].

Several studies have proposed an energy-efficient subway system to mitigate energy consumption of the subway. Sustainable energy management for the subway of the European research project was analyzed, and its energy system was evaluated, including life cycle assessment. The sustainable energy management system for underground stations reduced the total energy consumption of the target station by 13% and analyzed energy payback time [6]. An optimal energy model for the subway station was proposed by multiple regression considering the conditions of subway stations. The suggested model had the ability to decrease the energy required by more than 10% [7]. An energy assessment was implemented in a subway station and revealed the quantity of energy consumption, including that from the lighting system, ventilation process, signal antenna, and transportation system. This assessment suggested strategies to improve energy efficiency by use of alternative equipment and systems [3]. A multi-layered perceptron-based model was employed to estimate the intensity of energy consumption of subway station for optimizing mechanical and electrical systems [2]. However, the recent role of the subway station is not limited to an energy efficient transportation system. Providing a comfortable environment for passengers is also important. In addition, climate change has produced variation of meteorological conditions, such as temperature, precipitation, and concentration of outdoor particulate matter 10 μm or less in diameter (PM₁₀) [4]. Thus, climate change increases PM₁₀ concentration, the contaminated air penetrates and flows to the platform of the subway through the ventilation system [8]. This mechanism directly poses a health threat to passengers [9]. Additionally, excessive ventilation energy contributes to climate change, which may be increased to mitigate the PM₁₀ concentration in the subway station.

Therefore, to improve the ventilation system, several studies have been conducted to provide a healthy environment for passengers while decreasing energy consumption. Air quality monitoring with analyzing energy savings was conducted on subway systems with different operating systems [10]. A ventilation system with platform screen doors was suggested to decrease the operating energy by 20% and to improve the thermal environment [11]. A system for autonomous control of heating, ventilating, and air conditioning with environmental monitoring was deployed in a subway system using a linear regression model [12]. A three-dimensional computation model was developed to optimize the thermal environment while decreasing ventilation energy by regulating the train speed [13]. Although the suggested ventilation systems increased energy efficiency and environmental conditions, they have not been able to be used proactively because of the complexity of climate conditions. Extreme changes in weather and climate events can be occurred by climate changes or other disturbances [4]. And the changes can cause deterioration of outdoor air quality (OAQ) (i.e., outdoor PM₁₀), which can cause ventilation system malfunctions [14]. Unexpected and uncertain variation of OAQ can decrease the performance of ventilation systems; in this respect, a proactive ventilation system can be a useful alternative to prevent loss of ventilation performance.

Additionally, OAQ, which has the highest correlation value with IAQ, is the most influential factor to increase indoor PM₁₀ concentration, as shown in Fig. 1. OAQ and indoor air quality (IAQ) have similar trend on all operational times. And numbers of passengers and subways are related to the IAQ contamination during rush hours. Thus, OAQ has the highest correlation value among the three variables. More detail information of the correlations is described in Appendix A. These indicate the importance of prediction of OAQ to create a strategy not only to decrease indoor PM₁₀, but also to increase energy efficiency. Therefore, it is necessary to consider the complex variation of OAQ through a prediction model and to associate that with an optimal ventilation system in terms of economic and environmental improvements. Conventional ventilation control methods have been applied and suggested corresponding to the changing OAQ. The manual ventilation system fixed the ventilation flow rate and showed unclean IAQ to the passengers [4]. A model-based predictive control system was proposed to regulate ventilation frequency but there was no guarantee of optimal environment because of the short term operation optimization [15]. A gain-scheduling ventilation system scheduled a continuous ventilation response, however, it had the weakness to consider the stochastic behaviors by OAQ such seasonality and yellow storm events [16]. In addition, the inaccuracies of sensors have hindered the application of the real-time-dependent conventional systems. Therefore, there has been growing interest in the proactive system so as to proactively guarantee economic and environmental operation coping with the uncertainty of OAQ.

This study suggests a novel proactive energy-efficient and environmentally proactive optimal ventilation system under several OAQ conditions corresponding to climate change. First, a prediction model was run through a mode decomposition technique, and the model predicted the next 24-h OAQ and the number of passengers based on historical data. Two deep learning models, deep neural network (DNN) and gated recurrent unit (GRU), were employed, and the results were compared with those of a conventional statistical model, multiple linear regression (MLR). Additionally, this study evaluated prediction performances of deep learning models and the conventional model using statistical evaluators. Then, the predicted data were used to provide a trajectory of optimal ventilation flow rates for the next 24 h through artificial intelligence (AI)-iterative dynamic programming (IDP). The proposed proactive ventilation system was applied to the D-subway station in Korea by model-based simulation, and its economic and environmental benefits were compared with those of the manual operation system. The importance and contributions of this paper can be summarized as follows:

• Optimal energy-efficient operational strategies are considered for a ventilation system that routinely has been operated in fixed ventilation flow rates. OAQ conditions predicted by a deep learning technique are incorporated into the energy-efficient ventilation system. It is notable that these predicted OAQ conditions have not been considered in the previous ventilation systems. One day ahead OAQ conditions for optimizing the ventilation system are obtained through the deep learning technique. Different from previous studies, the proposed energy-efficient ventilation system is designed to search optimal operational strategies for the next 24 h instead of immediately controlling flow rate at the time. It can facilitate the ventilation system coping with un-expectable energy and IAQ problems under extreme OAQ conditions.

• A novel AI-IDP-based proactive ventilation system automatically provides optimal ventilation strategies considering economic and environmental benefits. This system does not require human intervention and the operator’s knowledge. Thus, it is quite different from most of the existing works on the exact domain knowledge-based systems such as the air quality monitoring system [10], the ventilation system with screen doors [11], the linear regression-based control system [12], and the three-dimensional computation model [13]. They require the knowledge of operators and are hard to predict and control unnoticed events. The proactive AI-IDP is applied to search the optimal future strategy, in the sense that a priori knowledge is not necessary, thus making the operational strategies more reliable and robust under unexpected OAQ conditions.

• The results show the effectiveness of the proposed proactive energy-efficient ventilation system. Evaluation of the proposed system is conducted considering consumed energy, operational costs, GHG emissions, and environmental health index.