RBF neural network modeling and application based on improved sparrow search algorithm

Using sparrow search algorithm (SSA) to optimize neural network model is easy to fall into local extremum, which leads to low accuracy of soft sensor. To solve this problem, this paper proposes an improved sparrow search algorithm (ISSA) to optimize the widths, centers and weights of radial basis function neural network (RBFNN). Taking the mean square error value as the individual fitness value, the position update formulas of discoverers, participants and early warning persons in SSA are improved to optimize the parameters of RBFNN. Finally, the convergence of the algorithm was proved, and the chemical oxygen demand (COD) was predicted using the random data validation and the measured data of the sewage treatment plant. The results showed that the mean absolute percentage error of SSA-RBF neural network was 1.55%, the mean absolute percentage error of ISSA-RBF neural network was 1.26%, and the mean absolute percentage error decreased by 0.29%, It shows that ISSA-RBF neural network has high soft-sensing accuracy.