基于PSO-CNN-LSTM的短期热负荷预测模型
Short-Time Heating Load Prediction Model Based on PSO-CNN-LSTM n
谢文举 1薛贵军 2白宇1
作者信息
- 1. 华北理工大学电气工程学院,河北 唐山 063210
- 2. 华北理工大学智能仪器厂,河北 唐山 063000
- 折叠
摘要
为提高短期供热负荷预测精度,减少供热不均与供需失调所造成的能源浪费,提出一种基于粒子群(Particle swarm op-timization,PSO)、经验模态分解(Empirical Mode Decomposition,EMD)、卷积神经网络(Convolutional Neural Network,CNN)和长短时记忆神经网络(Long Short Term Memory Network,LSTM)的混合预测模型.首先,针对供热负荷呈现非线性、复杂性等特点,采用EMD对供热负荷分解,从而实现弱化供热负荷复杂程度;其次,分别运用CNN与LSTM提取供热负荷空间特征与时域特征;最后,结合PSO算法对LSTM网络的超参数进行调整,寻找出最优参数.实验表明,结合EMD分解的PSO-CNN-LSTM网络相比LSTM、CNN-LSTM、EMD-CNN-LSTM平均误差分别降低了 44%、34%、24%、21%,拥有更高的预测精度和拟合效果.所提模型为集中供热负荷预测提供了一种新的思路,对于制定集中供热能源分配提供了参考意义.
Abstract
In order to improve the accuracy of short-term heating load prediction and reduce the energy waste caused by uneven heating supply and demand imbalance.In this paper,a hybrid prediction model based on Particle swarm optimization(PSO),Empirical Mode Decomposition(EMD),Convolutional Neural Network(CNN)and Long Short Term Memory Network(LSTM)is proposed.Firstly,for the characteristics of nonlinearity and complexity of the heating load,EMD was used to decompose the heating load so as to weaken the complexity of the heating load;Sec-ondly,CNN and LSTM were used to extract the spatial and time-domain features of the heating load respectively;Fi-nally,the hyperparameters of the LSTM network were adjusted in combination with PSO algorithm to find the optimal parameters.The experiments show that the PSO-CNN-LSTM network combined with EMD decomposition has 44%,34%,24%and 21%lower average error compared with LSTM,CNN-LSTM and EMD-CNN-LSTM,respectively,and possesses higher prediction accuracy and fitting effect.The proposed model provides a new way of thinking for central-ized heating load prediction and provides reference significance for developing centralized heating energy allocation.
关键词
碳中和/经验模态分解/粒子群优化/长短时记忆网络Key words
Carbon neutrality/Empirical mode decomposition/Particle swarm optimization/Long short term memory network引用本文复制引用
基金项目
国家自然科学基金青年基金(61502143)
出版年
2024
NETL
NSTL
万方数据