基于Apriori-CNN-LSTM的公共建筑供热系统用户侧阀门调控方法
User Side Valve Opening Prediction of Public Building Heating System Based on Apriori-CNN-LSTM
孙银兵 1李慧 1高祥宇1
作者信息
- 1. 山东建筑大学 热能工程学院,济南 250100
- 折叠
摘要
针对公共建筑供热系统中依赖人工经验调节用户阀门开度,无法实现智能调节造成人力资源浪费问题,该文提出了一种基于Apriori关联挖掘特征变量的CNN-LSTM混合神经网络用户阀门开度预测模型,用于阀门预测调控,让阀门调节变得更加智能、高效.该文首先从系统中导出前期人工调节供热系统运行的历史数据,在数据清洗后使用Apriori关联规则算法挖掘各变量数据与调节阀开度之间的关系,形成关联规则库,通过分析与调节阀开度相关的强关联规则选出特征变量.结合数据特点,选择结合卷积神经网络的CNN-LSTM混合模型来预测调节阀开度,计算误差对比发现预测结果优于单一的LSTM和Attention-LSTM神经网络模型.通过神经网络算法预测阀开度实现智能化调节,保证管网内支路流量与热量的动态平衡,满足用户对供暖舒适性的要求.
Abstract
A CNN-LSTM hybrid neural network user valve opening prediction model based on Apriori association min-ing feature variables is proposed to address the issue of human resource waste caused by relying on manual experi-ence to adjust user valve opening in public building heating systems,which makes valve adjustment more intelligent and efficient.This article first exports historical data of the manually adjusted heating system operation in the early stage from the system.After data cleaning,the Apriori association rule algorithm is used to mine the relationship be-tween variable data and the opening of the regulating valve,forming an association rule library.By analyzing the strong association rules related to the opening of the regulating valve,characteristic variables are selected.Based on the characteristics of the data,a CNN-LSTM hybrid model combined with convolutional neural networks was selected to predict the opening of the regulating valve.Comparison of calculation errors revealed that the prediction results were superior to the single LSTM and Attention-LSTM neural network models.By using neural network algorithms to predict valve opening and achieve intelligent regulation,the dynamic balance between branch flow and heat in the pipeline network is ensured,meeting the user's requirements for heating comfort.
关键词
供热系统/Apriori关联规则算法/LSTM神经网络/卷积神经网络/Attention注意力机制Key words
heating system/Apriori association rule algorithm/LSTM neural network/convolutional neural networks/Attention mechanism引用本文复制引用
出版年
2024
NETL
NSTL
万方数据