摘要
针对药品库中上层空间温度分布不均匀、环境温度信息采集中单传感器传输数据误差大、可靠性低可能导致实时监测效果不佳的问题,提出一种分层分簇无线传感器网络实时融合策略。该策略分为三层,其中在底层,利用改进的无迹卡尔曼滤波器对传感器采集的温度数据进行预处理;在中间层,每个簇头作为一个局部融合中心,通过并行快速对角阵权系数协方差交叉融合算法(PFDCI)对预处理的数据进行融合;在顶层的全局融合中心提出一种融合卷积神经网络和多策略天鹰算法优化改进极限学习机的网络结构处理局部融合数据,实现对药品库内环境温度的精准监测。仿真结果表明,该策略的局部融合算法与全局融合算法对测试样本的均方根误差、平均绝对误差和平均绝对百分比误差分别为0.026℃、0.020℃、0.11%和0.016℃、0.013℃、0.07%,完全满足现场测量需求。经验证,该融合策略在保证数据融合的稳定性和准确性的同时,减少了外部干扰,提高了药品库环境温度监测的精确度。
To address the problems,such as uneven temperature distribution in drug stores,large data transmission errors of single sensors in environmental temperature information collection,and low reliability which may lead to poor real-time monitoring.A real-time fusion strategy based on a layered and clustered wireless sensor network is proposed.The strategy is divided into three layers,where in the bottom layer,the temperature data collected by sensors are preprocessed using an improved traceless Kalman filter;in the middle layer,each cluster head acts as a local fusion center and the preprocessed data are fused by a parallel fast diagonal array power coefficient covariance cross-fusion algorithm(PFDCI);in the top layer,the global fusion center uses a fused convolutional neural network,and its structure is optimized by a multi-strategy Skyhawk algorithm to process the local fusion data and realize the accurate monitoring of the environmental temperature in the drug store.Simulation results show that for the test samples,the local fusion algorithm and the global fusion algorithm of this strategy have root mean square error,mean absolute error and mean absolute percentage error of 0.026℃,0.020℃,0.11%and 0.016℃,0.013℃,0.07%respectively,which fully satisfy the requirements of field measurements.It is verified that the fusion strategy ensures the stability and accuracy of data fusion while reducing the external interference and improves the accuracy of ambient temperature monitoring in the pharmaceutical warehouse.
作者
朱本科
高丙朋
蔡鑫
ZHU Benke;GAO Bingpeng;CAI Xin(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China)
出处
《中国测试》
北大核心
2025年第8期182-192,共11页
China Measurement & Test
基金
国家自然科学基金(62263031)
新疆维吾尔自治区自然科学基金(2022D01C694)
自治区高校基本科研业务费科研项目(XJEDU2023P025)。
关键词
药品库
温度监测
无线传感器网络
数据融合
无迹卡尔曼滤波
极限学习机
drugs warehouse
temperature monitoring
wireless sensor networks
data fusion
unscented Kalman filtering
extreme learning machine
