基于机器学习的集中供热负荷预测模型研究
Research on central heating load prediction models based on machine learning
张国正 1李化淼 1王志成 1张玉中 1陈君2
作者信息
- 1. 同方节能工程技术有限公司,北京 100084
- 2. 临沂市恒源热力集团有限公司,山东 临沂 276004
- 折叠
摘要
为保证用户的用热需求,提高能源利用率,精准预测供热负荷,将用户投诉率、室外气象条件以及历史负荷作为输入特征,利用BP神经网络、遗传算法优化BP神经网络(GA-BP)、支持向量回归机(SVR)和长短期记忆网络(LSTM)分别建立了 4种供热负荷预测模型,对未来一段时间的供热负荷进行逐时预测,并以实际供热数据进行训练和验证.研究结果表明:相较于其他 3种预测模型,GA-BP的预测精度最高,拟合优度R2 可达 0.994,平均绝对百分比误差最小为 1.20%,可应用于实际供热负荷预测与调控.
Abstract
In order to ensure the heat demand of customers,improve the energy utilization rate,and accurately predict the heating load.Using customer complaint rate,outdoor weather conditions and historical load data as input features,four load prediction models are developed by using BP neural network,genetic algorithm optimized BP neural network(GA-BP),support vector regression machine(SVR)and long short-term memory network(LSTM)to predict the heating load for a future period on a time-by-time basis,and are trained and validated with actual heating data.The results show that compared with the other three prediction models,GA-BP has the highest prediction accuracy,the goodness-of-fit R2 can reach 0.994,and the minimum average absolute percentage error is 1.20%,which can be applied to the actual heating load prediction and regulation.
关键词
供热负荷预测/BP神经网络/遗传算法/支持向量回归机/长短期记忆网络Key words
heating load prediction/BP neural network/genetic algorithm/support vector regression machine/long short-term memory network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据