Tue 23 Nov, 3:00, WeH 4601


	   The Application of Reinforcement Learning to the
	     Compensation of Reactive Power Disturbances


			    Mark Sheppard
		    School of Computing and Maths
			University of Teesside

			   mark@tees.ac.uk



An arc furnace operates by passing high voltage electricity through
metal within a crucible. As the system short circuits, the metal
melts. However as the metal progresses from its solid to molten state,
the load it places on the circuit is continually changing.

This varying load is inductive and can be compensated by the
introduction of an equal capacitive load. However existing measurement
techniques are less than ideal, since they are static and do not
adequately model the complexities of the system.

The use of reinforcement learning to control the compensation is being
considered.