Coke Ratio Prediction Based on Immune Particle Swarm Neural Networks

The clonal selection mechanism and vaccination strategy of immune system are introduced into particle swarm optimization algorithm in this paper, in order to enhance the ability of global exploration of PSO, avoiding getting into local optimum and improving the accuracy and convergence speed of BP networks. The global Cauchy mutation operator and local Gauss mutation operator are used to improve the ability of searching global optimization and the accuracy of local optimization. Then the weights and thresholds of neural networks are trained by applying the immune particle swarm optimization. Finally the coke ratio forecasting model is established based on the modified BP neural networks optimized by immune particle swarm optimizer. The result shows the forecast accuracy is more accurate than both the BP neural networks optimized by the standard PSO and the traditional BP neural networks, and provides an effective way to reduce the coke ratio and achieve energy conservation and emission reduction for iron and steel enterprise.