供稿: 庞恺, 丰励, 郑文超 | 时间: 2024-05-15 | 次数: |
庞恺, 丰励, 郑文超.基于微波辐射测量和GA-BP神经网络的电缆内部温度反演[J].河南理工大学学报(自然科学版),2024,43(3):146-153.
PANG K , FENG L , ZHENG W C .Inversion of cable internal temperature based on microwave radiation measurement and GA-BP neural network[J].Journal of Henan Polytechnic University(Natural Science) ,2024,43(3):146-153.
基于微波辐射测量和GA-BP神经网络的电缆内部温度反演
庞恺, 丰励, 郑文超
湖北工业大学 太阳能高效利用及储能运行控制湖北省重点实验室,湖北 武汉 430068
摘要: 目的 针对现有电缆监测系统无法实现无损式测量电缆内部温度的问题, 方法 提出一种基于微波辐射测量的电缆内部温度反演方法。首先,使用非相干方法构建多层微波辐射传输模型,利用指数函数进行分层,通过无限元仿真得到微波辐射计测量的亮温数据集,应用十折交叉验证划分数据集,为模拟实际测量中微波辐射计测得的亮温值易受环境噪声影响的情况,本文将测试集仿真值混合信噪比10~50 dB的高斯白噪声作为微波辐射计的实测值;然后采用遗传算法(genetic algorithm,GA)优化BP神经网络的权值和阈值,以此构建GA-BP神经网络模型,再向构建好的网络中输入样本,训练网络模型;最后利用实测值验证经过训练的GA-BP神经网络模型在反演电缆内部温度方面的能力。结果 实验结果表明:相较于仅使用BP神经网络,使用遗传算法优化后的BP神经网络在反演电缆内部温度方面表现更优异;网络的反演效果和实测值添加的噪声信噪比呈正相关,噪声信噪比越大,神经网络的反演效果越好;当测试集中添加的噪声信噪比大于36 dB时,R2 >0. 627,RMSE<5.55,反演效果较好;当混合噪声信噪比为50 dB时,反演效果最优,决定系数可达0.985。 结论 结果证明了GA-BP神经网络在电缆内部温度反演的有效性,为电缆内部温度反演提供了新思路。
关键词:电缆内部温度;非均匀分层;多层电缆微波辐射传输模型;BP神经网络;遗传算法
doi:10.16186/j.cnki.1673-9787.2022050029
基金项目:国家自然科学基金资助项目(41601399);太阳能高效利用及储能运行控制湖北省重点实验室开放基金项目(HBSEES202007);湖北工业大学博士科研启动基金资助项目(BSQD2020011)
收稿日期:2022/05/10
修回日期:2022/10/05
出版日期:2024/05/15
Inversion of cable internal temperature based on microwave radiation measurement and GA-BP neural network
PANG Kai, FENG Li, ZHENG Wenchao
Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System,Hubei University of Technology,Wuhan 430068,Hubei,China
Abstract: Objectives The current cable monitoring system could not measure the cable internal temperature non-destructively. Methods Therefore,based on microwave radiation measurement,a cable internal temperature inversion method was proposed.Firstly,the multilayer microwave radiation transmission model was constructed by using the incoherent approach,and the exponential function was used for layering.The bright temperature data set measured by microwave radiometer was obtained through finite element simulation.The 10-fold cross-validation was applied to divide the data set.To simulate the effect of the environmental noise on microwave radiometer measured brightness temperature in actual measurement,Gaussian noise with SNR ranging from 10 ~ 50 dB was added to the calculated value as the measured value for the microwave radiometer.After that,the GA-BP neural network model was constructed,and then the network was trained by using the data set.Finally,the inversion performance of the pre-trained GA-BP neural network model was verified by using the measured values. Results The specific conclusions were as follows:Compared with the one that used the BP neural network only,the optimized BP one adopting a genetic algorithm was more effective in the cable internal temperature inversion.The inversion effect of the network and the signal-to-noise ratio (SNR) of added noise were positively correlated.The larger the SNR,the better the inversion effect.When the SNR was greater than 36 dB,the coefficient of determination was greater than 0. 627,and the root mean square error was less than 5. 55,which indicated that the inversion performed well.The inversion performance probably attained best when the SNR was 50 dB.The coefficient of determination could reach 0. 985. Conclusions The results proved that GA-BP neural network was effective for cable internal temperature inversion,lending a new perspective for cable internal temperature inversion.
Key words:internal cable temperature;non-uniform stratification;multilayer cable microwave radiation t-ransmission model;BP neural network;genetic algorithm