基于支持向量机的水质浊度补偿研究

Turbidity sensor compensation method for an SVM based on particle swarm algorithm

摘要:

针对传统浊度传感器的非线性误差,无法满足直接对水中浊度进行精确测量的需求,提出了一种支持向量机的方法补偿其性能。而支持向量机中惩罚系数和核参数决定了其补偿的性能,传统支持向量机寻参方法速度慢、运算量大,具有一定的局限性。针对其参数的选择优化提出了改进的网格搜索法优化支持向量机,即采用改进的网格搜索法来针对水质浊度监测传感器补偿系统的特性来优化选择惩罚系数和核参数。实验结果表明,基于网格搜索法的支持向量机测量精度达到93.0%,其各项测量误差满足实际标准要求。

Abstract:

Aiming at the non-linear error of the traditional turbidity sensor, which can"t meet the demand of directly measuring the turbidity in water, a support vector machine method is proposed to compensate its performance. The performance of compensation is determined by the penalty coefficient and kernel parameter in SVM. The traditional method of finding parameters in SVM is slow and requires a lot of computation, which has some limitations. An improved grid search method is proposed to optimize support vector machine (SVM) for the selection and optimization of its parameters. That is to say, the improved grid search method is used to optimize the selection and compensation of water quality turbidity monitoring sensor compensation system. The experimental results show that the measurement accuracy of SVM based on grid search method is 93.0%, and the measurement errors meet the actual standard requirements.