基于深度CRBM模型的建筑能耗预测方法  

Building Energy Consumption Forecasting Method Based on Depth CRBM Model

作　　者：李鹏 周希霖 

机构地区：[1]新疆铁道勘察设计院有限公司建筑分院,新疆乌鲁木齐830011 [2]日本东北大学大学院,日本仙台〒980-8579

出　　处：《湘潭大学自然科学学报》2017年第2期45-48,70,共5页Natural Science Journal of Xiangtan University

基　　金：国家自然科学基金项目(51408303);国家留学基金项目(201606950013)

摘　　要：针对建筑能耗的预测问题,提出一种基于深度条件受限玻尔兹曼机(CRBM)的预测方法.首先,将传统受限玻尔兹曼机进行扩展,融入一个历史条件输入层,使其能够根据历史时间序列来预测未来序列.然后,在CRBM基础上构建深度CRBM模型,用来执行建筑能耗的预测.在一个"个体家庭电力消耗"数据集上的实验结果表明,提出的方法能够准确预测出预定时间段内的建筑能耗,能够为电力调度提供一定的依据.For the issue that the prediction of building energy consumption,aprediction method based on deep conditional limited Boltzmann machine（CRBM）is proposed.First,the traditional Restricted Boltzmann Machine is extended to a historical condition input layer,so that it can be based on historical time series to predict the future sequence.Then,a depth CRBM model is built on the basis of CRBM to perform the prediction of building energy consumption.The experimental results on an ＂individual household power consumption＂ dataset show that the proposed method can accurately predict the building energy consumption for a predetermined period of time and can provide some basis for power dispatching.

关 键 词：建筑能耗预测 深度条件受限玻尔兹曼机 历史条件输入层 时间序列 

分 类 号：TU111[建筑科学—建筑理论]