化工企业烧碱产能预测的研究

Prediction of Caustic Soda Production Capacity in Chemical Industry

为了有效的提高化工企业烧碱产能预测的精度,针对影响烧碱产能多因素间的不确定性和因素间的交互作用,提出了一种施密特正交马田系统(MTGS)与粒子群算法优化BP神经网络(PSO_BP)相结合的烧碱产能预测方法。采用施密特正交马田系统把影响烧碱产能的七个因素减少为六个核心因素,减少了PSO_BP的输入量,提高了烧碱产能预测系统的快速性;对粒子群算法优化的BP神经网络建模,构建MTGS_PSO_BP的烧碱产能预测模型,提高了预测的精度。最后用MATLAB进行仿真测试,与BP神经网络的预测结果相比,预测误差明显小于BP神经网络,具有更高的预测精度,为烧碱产能预测提供了一种新的科学、有效的方法。

In order to improve the accuracy of caustic soda production in chemical enterprises, we put forward a new MTGS and PSO_BP mixed prediction algorithm aiming at the uncertainty and interaction of the factors that can influence the production of caustic soda. The Schmidt Orthogonal Martin System reduces the seven influencing factors of caustic soda capacity to the six core factors of them. This change reduces the input of the PSO_BP system and im- proves the rapidity of prediction system of caustic soda production. We establish a model for BP neural network which is optimized by the Particle Swarm Optimization algorithm and build the MTGS_PSO_BP production prediction model that improves the prediction accuracy. At last, a simulation testing is done by MATLAB. Simulation results show that, compared with prediction results of BP neural network, the new model has a higher prediction precision.