Loading...

Operational Planning & Control of Multi-Vector Energy Networks Using Distributed Intelligent Methods

Khani, Arman | 2023

100 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55805 (06)
  4. University: Sharif University of Technology
  5. Department: Chemical and Petroleum Engineering
  6. Advisor(s): Bozorgmehry Boozarjomehry, Ramin
  7. Abstract:
  8. Multi-vector energy (MVE) networks are one of the most complex systems which are attractive to many researchers. In this work, an integrated gas-power transmission network as an MVE network has been studied. The MVE benchmark used in this work comprises a gas transmission network, which is the core of the MVE, a power transmission network, and a Syn-Gas plant. The gas transmission network supplies the gas demands of its consumers. The main objective of this thesis is to design appropriate controllers to maintain the states of the gas network in its desired condition. Since the gas network is distributed parameter system and its model is an infinite dimensional state space, it is either impossible or very computationally demanding to use model-based controllers for this type of systems. Therefore, a set of intelligent controllers is used to achieve this work's control objectives. These objectives are to supply the demands of the consumers via minimum consumed energy while the pressure of various points in the gas network lies in its standard operating range. The gas transmission network has been simulated using Python programming language to achieve the mentioned goal and evaluated in an open-loop simulation. The controllers whose manipulating variables are compression ratios of various compress stations are designed and implemented based on reinforcement learning according to the actor-critic approach, mainly because their input-output data are not known beforehand. Furthermore, all controllers use the same actor and critic agents but their inputs comprise of both dynamic and static variables defined based on the local environments of each controller and those of the global environment (i.e., the whole gas network). This approach provides the benefit of generalizability and makes it applicable in various gas transmission networks regardless of their topology and design. The performance of the designed controllers has been evaluated based on the transient behavior of the gas network. The obtained results show the designed controllers are capable of handling both regulatory and supervisory modes
  9. Keywords:
  10. Gas Transmission Network ; Reinforcement Learning ; Intelligent Control ; Complex System ; Integrated System ; Multi-Vector Energy

 Digital Object List

 Bookmark

No TOC