Write your presentation abstract here and put the paragraph delimiter at the end of paragraph.
Industrial process control is a very important topic as it is both the basis and embodiment of today's technology and science.
However, controlling a large-scale process could be very complicated and expensive in terms of required methods and resources. Traditional methods
entail a complex model from where we unavoidably suffer the modelling error.
This master research explores a model-free Reinforcement Learning (RL) method for controlling a larg-scale chemical process: Vinyl Acetate Monomer process
which is very important to today's chemical industry. Our RL method does not require model knowledge and learn to control the plant via trial-and-error
interaction with the environment (plant) from scratch.
The main contribution of this master research is that, it managed the difficulties from process control problems e.g., high dimensionality of samples.
The high dimensionality is a result of continuous-valued readings of many sensors that consitituent the inner engineering structure of a manufacturing process.
Specifically, two novel RL algorithms are proposed in this master research to handle two different kinds of problems one will usually encounter in
the process control domain, i.e., local control and plant-wide control. Experiments of the two algorithms show comparable performance to the state-of-the-art model-based controllers given some assumptions hold.