摘要
为了使公共建筑辐射空调更加节能,需合理评估空调能耗变化,为此,提出公共建筑辐射空调能耗机器学习动态评价模型。机器学习的空调能耗动态评价模型,采用基于均值漂移聚类的数据清洗算法剔除异常数据,保留正常数据,并将其输入KNN填充算法,补充空调能耗数据集中缺失的数据点,使数据更加完整;将处理完的有效数据输入Adaboost-BP算法,不断迭代后,分类空调能耗预测数据,动态评价公共建筑辐射空调能耗。经实验验证,该模型在预测空调用电量时相对误差保持在1%及以下,可以实现精准的用电量分析,还可有效预测不同季度下的空调热负荷、天然气总消耗、二氧化碳排放量等多种能耗,实现公共建筑辐射空调的合理评价。
In order to make radiant air conditioning in public buildings more energy-saving,it is necessary to reasonably evaluate the change of air conditioning energy consumption.Therefore,a machine learning dynamic evaluation model of energy consumption for radiant air conditioning in public buildings is proposed.This paper designs a dynamic evaluation model of air conditioning energy consumption based on machine learning,and uses a data cleaning algorithm based on mean shift clustering to eliminate abnormal data,retain normal data,input them into KNN filling algorithm,and supplement the missing data points in the air conditioning energy consumption data set to make the data more complete.The processed effective data are transmitted to the Adaboost-BP algorithm.After continuous iteration,the predicted data of air conditioning energy consumption are classified to dynamically evaluate the radiant air conditioning energy consumption in public buildings.Experiments show that the relative error of this model in predicting air conditioning power consumption is kept at 1%and below,which can achieve accurate power consumption analysis,and can effectively predict air conditioning heat load in different seasons,total natural gas consumption,carbon dioxide emissions and other energy consumption,so as to achieve a reasonable evaluation of radiant air conditioning in public buildings.
作者
巨健
俞文瑾
曹敏
尹璐
弋欣
JU Jian;YU Wenjin;CAO Min;YIN Lu;YI Xin(State Grid Shaanxi Electric Power Co.,Ltd.,Xi’an 710051,China;Marketing Service Center of State Grid Shaanxi Electric Power Co.,Ltd.(Metrology Center),Xi’an 710051,China)
出处
《微型电脑应用》
2024年第12期174-178,共5页
Microcomputer Applications
关键词
公共建筑
辐射空调
机器学习
动态评价
评价模型
public building
radiant air conditioning
machine learning
dynamic evaluation
evaluation model
作者简介
巨健(1986-),男,本科,高级工程师,研究方向为电力营销、电能替代;俞文瑾(1989-),男,本科,工程师,研究方向为电力营销、电能替代、能效管理;曹敏(1980-),男,硕士,高级工程师,研究方向为电力营销、电力需求侧管理、综合能源、市场化交易;尹璐(1990-),女,本科,研究方向为电力营销;弋欣(1990-),女,蒙古族,本科,工程师,研究方向为电力营销、市场化交易。
