Today I had a supervisor meeting with Alex and Gopal. In it I explained my simple decentralised message passing algorithm for determining which generators need to ramp up in order to reduce an overloaded line. As in the scenario I described in my previous blog post. However Alex pointed out that the way in which I allocate power does not take into account when a load overloads a line but the load is not directly connected to the line that gets overloaded. In this case a node nearer the load may be able to satisfy it however in my model I had assumed that load was situated at the root of the tree.
Therefore I have turned my attention to solving the decentralised power flow in a tree using dynamic programming. Is there an allocation of generator outputs that satisfy the thermal limits of the transmission lines and satisfy the loads within the tree using a decentralised dynamic programming algorithm. As a starting point I am going to read the following paper which does a similar thing for sending information around a tree using sensors:
Kho J, Tran-Thanh L, Rogers A, Jennings NR. An Agent-Based Distributed Coordination Mechanism for Wireless Visual Sensor Nodes Using Dynamic Programming. The Computer Journal. 2010;53(8):1277.
Initially from the dynamic programming explanations I think each generator will represent a stage and the state of each stage will be a continuous variable representing the generators output. The decision variable for each stage will represent the optimal output for each generator such that carbon emissions are minimised, the flow within the network is satisfied and the transmission lines are not overloaded.
No comments:
Post a Comment