Ana Soares from the University of Coimbra presented her work on "Domestic Load Scheduling Using Genetic Algorithms" where a Genetic Algorithm is used to optimize for an objective function considering energy consumption, end user preferences, peak power, and presently available energy. Encoding of solutions was done as string of integers where the recombination was done by a bit mask over the integer string (so no typical crossover). The evolved results define scheduling of loads from household appliances in order to fulfill the above defined objectives.

Stephan Hutterer from FH Hagenberg approached the optimal power flow problem with an evolutionary algorithm. Optimal control policies are learned offline for a given power grid resulting in general abstract rules for optimal power flow.

*"Prediction is difficult, especially of the future*" (Nils Bohr) - the prediction of power load profiles can be improved with the approach presented by Frédéric Krüger from the Université de Strasbourg. They show how a genetic algorithm generated with the EAsy Specification of Evolutionary Algorithms (EASEA) language can be applied to solve a noisy blind source separation problem and create accurate power load profiles using real world data.

Another approach for forecasting electrical consumption was presented by Martina Friese and Oliver Flasch from FH Köln in his talk on "Comparing Ensemble-Based Forecasting Methods for Smart-Metering Data". They apply state-of-the-art time-series forecasting methods to electrical energy consumption data recorded by smart meters and show that genetic programming is an attractive alternative to custom-built approaches for electrical energy consumption forecasting.

Dominik Egarter from Alpen-Adria-Universität Klagenfurt presented the paper "Evolving Non-Intrusive Load Monitoring" [PDF]. Here, an evolutionary algorithm is used to determine a set of devices for a given load curve - in other words, your smart meter knows what devices you have on even if they are not smart. The work on evolving non-intrusive load monitoring shows the capabilities of the approach but also its limits. The latter basically tell you how much you have to masquerade your power profile so that it does not give away information about the devices that constituted it. See also this blog article on Dominik's work.

What about a Pareto based approach? .... like in this paper: http://www.mdpi.com/1996-1073/6/3/1439/pdf

ReplyDelete