Load forecasting in a smart grid through customer behaviour learning using L1-regularized continuous conditional random fields
RIS ID
112843
Abstract
Copyright 2016, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.Load forecasting plays a critical role in Smart Grid. As there have been various types of customers with different behaviours in a Smart Grid, it would benefit load forecasting if customer behaviours were taken into consideration. This paper proposes a novel load forecasting method that efficiently explores customers' power consumption behaviours through learning. Our method uses L1-CCRF to initially learn the behaviour of each customer, followed by a hierarchical clustering process to cluster all the customers according to their different behaviour patterns, and then fine-tunes a corresponding L1-CCRF to predict the load for each customer cluster, and finally, sums all the predicted loads of customer clusters to obtain the load for the whole Smart Grid. The proposed method utilizes L1-CCRFs to effectively capture the relationships between various customers' loads and a range of outside influential factors. Experiments from different perspectives demonstrate the advantages of our load forecasting method through customer behaviour learning.
Grant Number
ARC/DP140100974
Publication Details
Wang, X., Zhang, M. & Ren, F. (2016). Load forecasting in a smart grid through customer behaviour learning using L1-regularized continuous conditional random fields. AAMAS '16: International Joint Conference on Autonomous Agents and Multiagent Systems (pp. 817-825). ACM Digital Library: ACM.