A Study on Artificial Intelligent System considering the Predictive Model of Subway Station Air Quality

다국어 초록 (Multilingual Abstract)

Nowadays, air quality problems in subway stations are increasingly serious. Fine particles, such as PM10, PM2.5 are well known air pollutants which can cause many diseases. It is important to find solutions to reduce air pollution in subway station. Seoul subway which is used by more than four million people is facing severe air pollution. The Seoul Metro of Korea keeps making effort to reduce air pollution in subway stations and tunnel for people who use subway as daily transportation.
The aim of this thesis is the study on developing a framework design of Integrated Air Quality Management System as an Artificial Intelligent (AI) System which is utilized to improve air quality in subway station.
Thus, Seoul subway station was chosen as an experimental station in this thesis. The air quality in a subway station can be automatically monitored and controlled through this system. The measurements were conducted in four different compartments in a station so that the particle pollution sources and processes can be found. From these operations of Information Technology and Energy Technology (IT‐ET) and Artificial Intelligent (AI) system, the cleaner‐but‐greener air quality management system ran smoothly as expected. The experimental site is the subway station of Daecheong of line 3 with a large waiting room and eight exits. The national environmental regulation is that the 24‐hour average PM10 concentration should be lower than 150μg/m3. On the basis of the environmental regulation, Artificial Intelligent (AI) system on waiting room and platform ventilators were operated by only PM10 feedback controller data to control the ventilators. Then a predictive model on the AI system has been created to improve the ventilator efficiency and reduce the power consumption. The predictive model is composed of the real time data algorithm, predictive data algorithm and factors etc. At last this Artificial Intelligent (AI) system has been tested and verified, and it functioned well.
When the air quality was only controlled by feedback controller, the air quality was not controlled accurately on rush hour time. When the air quality was only controlled by feed forward controller, the air quality level would be high, because it was controlled per 30 minutes.
The AI system is consisted of feedback and feed forward controller which integrate both advantages and also eliminate their disadvantages. So the best air quality level and the lowest frequency level would be kept by AI system since the power consumption would be reduced by 20% on the rush hour time by AI system (Table 4-6). Once the PM10 data of another station is collected, analyzed and written the data in predictive model, the AI system can suit to every station.

목차 (Table of Contents)

• Chapter 1. Introduction = 1
• 1.1. Background = 1
• 1.2. Objective = 1
• Chapter 2. Equipments of Artificial Intelligence (AI) System = 3
• 2.1. Sensor Setting = 3
• 2.2. Setting the Communication Apparatus = 5
• Chapter 3. The Current Operating Condition of Ventilation Controller = 6
• 3.1. The Current Operating Condition = 6
• 3.2. The New Rule of Operating Condition = 7
• Chapter 4. Artificial Intelligence (AI) System = 9
• 4.1. The Objective of the Predicted Model Measurement = 9
• 4.2. The Condition of Measurement Local = 9
• 4.3. The Control Logic of AI System = 10
• 4.3.1. Feedback control = 10
• 4.3.2. Feed forward control = 11
• 4.4. The Logic Language = 12
• 4.5. Predicted Model Test and Verify = 15
• 4.5.1. The Testing and Verifying of Predicted Model Base Data = 15
• 4.5.2. The Benchmark of Judgment = 16
• 4.5.3. The Test and Verify of Variable Coefficient (Factor) = 18
• 4.5.4. The Result of (Factor) Measurement of Waiting Room = 24
• 4.5.5. The Result of (Factor) Measurement of Platform = 30
• 4.6. Comparing the Feedback Controller and Feed Forward Controller with Predicted Model = 31
• 4.6.1. The Result of Waiting Room = 33
• 4.6.2. The Result of Platform = 34
• 4.6.3. The Result of Comparison Feedback Controller and Feed Forward Controller of Predicted Model = 34
• Chapter 5. Conclusion and Future Work = 36
• 5.1. Conclusion = 36
• 5.2. Future Work = 37
• References = 39
• Abstract (in Korean) = 42