基于PSO优化BP神经网络的多传感器数据融合

Multi-sensor data fusion based on PSO optimized BP neural network

针对室内环境监测中单一传感器测量数据精度低、可靠性差的问题,提出一种基于粒子群(particle swarm optimization,PSO)优化BP神经网络多传感器数据融合算法。首先使用防脉冲干扰平均滤波算法来消除检测数据中的异常数据和噪声数据。其次,利用卡尔曼滤波算法对多同类传感器进行数据级融合,有效地降低因噪声干扰导致的测量误差,为异质传感器进行决策级融合提供最佳数据。最后,采用PSO优化BP神经网络算法进行决策级融合。实验结果表明,基于PSO优化BP神经网络多传感器数据融合算法对测试样本的平均绝对百分比误差(mean absolute percentage error,MAPE)和拟合度(r^(2))均优于BP神经网络和自适应加权(adaptive weighted,AW)优化BP神经网络,且运行时间比BP神经网络以及AW-BP神经网络分别短69.31%、50.36%。经验证,基于PSO优化BP神经网络多传感器数据融合算法具有更高的融合精度,同时缩短了算法的运行时间。

Aiming at the problems of low precision and poor reliability of single sensor measurement data in indoor environmental monitoring,a multi-sensor data fusion algorithm based on particle swarm optimization(PSO)optimization was proposed.Firstly,the anti-pulse interference averaging filtering algorithm is used to eliminate abnormal data and noise data in the detected data.Secondly,the Kalman filter algorithm is used to perform data-level fusion of multiple similar sensors,which can effectively reduce the measurement error caused by noise interference and provide the best data for decision-level fusion of heterogeneous sensors.Finally,the PSO optimization BP neural network algorithm is used for decision-level fusion.The experimental results show that the mean absolute percentage error(MAPE)and fit(r^(2))of the test samples based on the PSO optimized BP neural network multi-sensor data fusion algorithm are better than those of the BP neural network and adaptive weighed(AW)optimizes the BP neural network,and the running time is 69.31%and 50.36%shorter than the BP neural network and the AW-BP neural network,respectively.It has been verified that the multi-sensor data fusion algorithm based on PSO optimized BP neural network has higher fusion accuracy and shortens the running time of the algorithm.