Water Quality Sensor Model Based on an Optimization Method of RBF Neural Network

摘要

In order to solve the problem that the traditional radial basis function (RBF) neural network is easy to fall into local optimal and slow training speed in the data fusion of multi water quality sensors, an optimization method of RBF neural network based on improved cuckoo search (ICS) was proposed. The method uses RBF neural network to construct a fusion model for multiple water quality sensor data. RBF network can seek the best compromise between complexity and learning ability, and relatively few parameters need to be set. By using ICS algorithm to find the best network parameters of RBF network, the obtained network model can realize the non-linear mapping between input and output of data sample. The data fusion processing experiment was carried out based on the data released by Zhejiang province surface water quality automatic monitoring data system from March to April 2018. Compared with the traditional BP neural network, the experimental results show that the RBF neural network based on gradient descent (GD) and genetic algorithm (GA), the new method proposed in this paper can effectively fuse the water quality data and obtain higher classification accuracy of water quality.