摘要
有效预测电能负荷,对提高电力负荷时间序列测量准确度及合理制定用电能管理措施具有重要意义。针对传统预测模型在电能负荷预测中无法充分挖掘时间序列数据中隐藏特征的问题,基于电能数据时间序列的趋势,融合数值信息提出一种卷积神经网络(convolutional neuralnetwork,CNN)与长期短期记忆循环神经网络(long short-term memory network,LSTM)相结合的混合多隐层CNN-LSTM电力能耗预测模型。首先,通过设定最小目标函数作为优化目标,Adam优化算法更新神经网络的权重,并对网络层和批大小进行自适应调优以确定最佳层数和批大小。其次,构建混合多隐层模型并进行隐层组合优化与讨论,确定最佳时间维度的参数,进行时间维度的特征学习进而预测下一时间序列的耗电量。然后以某公司的电力负荷数据为例进行验证,并与LSTM、CNN、RNN等模型的预测结果分析比较。结果表明上述混合多隐层模型预测准确度达98.94%,平均绝对误差(MAE)达到0.0066,均优于其他相关模型,证明以上混合预测模型在电力负荷预测精度方面具有更好的性能。基于上述理论,开发了能耗监控决策系统,实现设备状态实时监控和能耗智能预测功能,为解决传统制造业能耗需求不精确和能源库存浪费问题提供参考和指导。
Effective prediction of electric energy consumption is important for improving the accuracy of electric load time series measurement and formulating reasonable measures for electric energy consumption management.To address the problem that traditional prediction models cannot adequately capture the evolutionary patterns of time variables in electric energy load prediction,this paper proposes a hybrid muli-hidden layer CNN-LSTM electric energy data time series based on the trend of electric energy data and fusing numerical information to propose a hybrid convolutional neural network and a long-term short-term memory recurrent neural network combined with a hybrid multihidden layer CNN-LSTM power energy prediction model.First,by setting the minimum objective function as the optimization objective,the Adam optimization algorithm updates the weights of the neural network and adaptively tunes the network layers and batch sizes to determine the optimal number of layers and batch sizes.Next,a hybrid multi-hidden layer model is constructed and the combination of hidden layers is optimized and discussed to determine the parameters of the optimal time dimension,and feature learning of the time dimension is performed to predict the power consumption of the next time series.Then the data set of a company's electricity load data is validated and compared with the prediction results of LSTM,CNN,RNN and other models.The results show that the prediction accuracy of this hybrid multi-hidden layer model can reach 98.94% and the mean absolute error(MAE)reaches 0.0066,both of which are better than other basic models,proving that this combined prediction model has better performance in terms of power load prediction accuracy.Based on the above theory,an energy consumption intelligent monitoring system is developed to realize real-time equipment status monitoring and energy consumption intelligent prediction function,which provides reference and guidance to solve the problems of inaccurate energy demand and energy inventory waste in traditional manufacturing industry.
作者
龚立雄
钞寅康
黄霄
陈佳霖
GONG Li-xiong;CHAO Yin-kang;HUANG Xiao;CHEN Jia-lin(School of Mechanical Engineering,Hubei University of Technology,Wuhan Hubei 430068,China)
出处
《计算机仿真》
2024年第8期77-83,共7页
Computer Simulation
基金
国家自然科学基金资助项目(51907055)。
关键词
电力负荷预测
卷积神经网络
长短期记忆神经网络
混合多隐层组合模型
Electric load forecasting
Convolutional neural network
Long and short-term memory neural network
Hybrid multi-hidden layer combination model
