摘要
区域供冷供热系统(DHC)负荷预测的精准度是实现系统设计合理和运行节能的关键因素。以某区域能源站实测数据为基础,采用多元线性回归(MLR)、分类与回归树(CART)和随机森林(RF)三种数据挖掘算法,分别建立了系统负荷时序预测模型,并对比了三种模型在不同时序变量影响下的负荷预测结果以及预测效率。结果表明:在不同时序变量影响组合下,随机森林模型的预测效果最好;在仅考虑预测精度的情况下,负荷的时序变量组合维度越复杂,预测结果精度越高,而预测效率越低;在同时考虑预测精度和预测效率的情况下,当时序变量采用室外干球温度、室外相对湿度、负荷(t-1)和负荷(t-2)的组合2时,随机森林模型的综合性能最好。
Accurate load forecasting is the key to achieve reasonable and more energy-efficient system of district heating and cooling(DHC).After collecting and analyzing historical load data of a regional energy station,three time-series prediction models of system load using are established by three data mining algorithms including multiple linear regression(MLR),classification and regression tree(CART)and random forest(RF),respectively.Meanwhile,the prediction results and prediction efficiency of the system load are compared using the three different models with different combinations of time-series variables.Results show that:(1)The prediction model based on the random forest model has the best prediction accuracy;(2)If only the prediction accuracy is considered,the more complex the dimension of time series variables combination for system load prediction,the higher the prediction accuracy and the lower the prediction efficiency;(3)If both prediction accuracy and prediction efficiency are considered,the random forest model obtained the optimal overall prediction performance when the time-series variables are combined into combination 2.
作者
焦良珍
陈海生
高革
李冠男
胡云鹏
JIAO Liang-zhen;CHEN Hai-sheng;GAO Ge;LI Guan-nan;HU Yun-peng(School of Architecture and Civil Engineering,Wuhan University of Science and Technology,Wuhan 430065,China;CEC Guanggu Architectural Design Institute Co.,Ltd.,Wuhan 430072,China;Wuhan Business School,Wuhan 430056,China)
出处
《建筑节能》
CAS
2020年第11期38-44,共7页
BUILDING ENERGY EFFICIENCY
基金
青年科学基金(51906181)
湖北省高等学校优秀中青年科技创新团队计划(T201829)
关键词
数据挖掘
时序分析
区域供冷供热
负荷预测
预测效率
data mining
time-series analysis
district heating and cooling
load forecasting
predictive efficiency
作者简介
焦良珍(1969),女,湖北人,环境工程专业,硕士,副教授,研究方向为空调系统节能及群控策略(656761762@qq.com)。
