Time Series Prediction Model Of B-Spline Network Based On Structural Parameters Optimization And Its Industrial Application

Time series prediction problems are involved in meteorology,economy,industrial production and other areas.Therefore,it is of great research significance to reasearch modeling and forecasting of the time series.In practical applications,many time series show very strong nonlinear characteristics,and the traditional time series prediction methods often fail to achieve the desired results.Therefore,it is an effective method to trect time series as nonlinear.Artificial neural network is often used to construct the prediction model based on nonlinear time series.And B-spline network is a kind of fuzzy associative memory neural network.Compared with other methods,the structure of B-spline network is simple,the parameters to be optimized are less,and the learning speed is fast.For this purpose,a nonlinear time series prediction model of B-spline network is proposed on the basis of structural parameters optimization of the network in this paper.Based on a background of the time series prediction,this paper mainly studies the construction of the prediction model of time series B-spline network,the optimization method of network structure parameters and its application in industrial production.The main research work is as follows:(1)Based on the depth analysis of the properties of the B-spline function and the studying of the structure and the operation mechanism of B-spline network,a new prediction model of B-spline network based on time series is constructed.In the design of network structure,the positions of nodes and the weight parameters of each spline basis function are together considered as independent variables to be optimized.(2)According to the selected evaluation function of the network,a parameter progressive search algorithm is designed to optimize the structure of the B-spline network.The validity of the prediction model is verified by the sinusoidal function and quadratic function.(3)An improved particle swarm optimization algorithm is proposed for structural parameter optimization of the B-spline network.The result of function verification of this algorithm is compared with the result of parameter progressive search algorithm.(4)The two methods are used to predict the quality of raw slurry in alumina production.The predicted results are compared with the radial basis function network.The results of prediction and comparison show that the B-spline network prediction model optimized by particle swarm algorithm has better prediction accuracy and generalization performance.The effect of its industrial application is better and it has strong practical value.It is an effective prediction method based on nonlinear time series.