Detecting Patterns of Appliances from Total Load Data Using a Dynamic Programming Approach - M. Baranksi, J. Voss
@inproceedings{Baranski:2004:DPA:1032649.1033472,
author = {Baranski, Michael and Voss, Jurgen},
title = {Detecting Patterns of Appliances from Total Load Data Using a Dynamic Programming Approach},
booktitle = {Proceedings of the Fourth IEEE International Conference on Data Mining},
series = {ICDM '04},
year = {2004},
isbn = {0-7695-2142-8},
pages = {327--330},
numpages = {4},
url = {http://dl.acm.org/citation.cfm?id=1032649.1033472},
acmid = {1033472},
publisher = {IEEE Computer Society},
address = {Washington, DC, USA},
}
Key Points
- Single sensor on the electricity meter
- Similar changes in electricity usage were clustered together and used to create FSM of appliances
- FSMs were optimised using shortest path first optimisation to ensure that each state change is optimally placed
Short paper describing a non-intrusive application load monitoring (NIALM) algorithm. Essentially each change in electricity load was detected, and similar changes were grouped together and said to describe a state change in an arbitrary state machine. Then, using genetic algorithms the search space of all combinations of state changes was performed to create a number of finite state machines (FSMs) which were a close fit to the system. These FSMs were then optimised further using a shortest path optimisation algorithm which was applied to each machine, and then globally.