Deep Neuroevolution for Energy Consumption Prediction

カテゴリ: 部門大会

論文No: OS5-9

グループ名: 【C】平成30年電気学会電子・情報・システム部門大会プログラム

発行日: 2018/09/05

タイトル(英語): Deep Neuroevolution for Energy Consumption Prediction

著者名: Irvan Mhd(東京工業大学)

著者名(英語): Mhd Irvan()

キーワード: Deep Learning|Machine Learning|Neural Networks|Genetic Algorithms|Renewable EnergyEnergy Prediction

要約(日本語): Deep Neuroevolution is a machine learning technique that combines the ideas from Deep Neural Networks (DNN) and Genetic Algorithms (GA). Recent advances in deep learning research showed that GA is a good alternative to train DNN. This is particularly applicable in deep reinforcement learning problems that requires predicting future events. Predicting consumers' future energy consumption fits into this area. Every person has a unique lifestyle which lead to various energy requirements at different time. It is challenging to predict how much energy they will consume in the next hours, days or weeks. In this study, we demonstrate how GA operations can be applied to evolve DNN parameters, such as their connections and weights, that can generate accurate prediction for this case.

PDFファイルサイズ: 191 Kバイト