基于模拟退火PSO-BP算法的钢铁生产能耗预测研究

针对BP对能源系统进行建模和预测的方法存在的问题,提出一种优化BP神经网络的模拟退火粒子群混合算法(SAPSO)。利用该混合算法优化BP神经网络的权值和阈值,然后训练BP神经网络预测模型以得到最优解,并将所建立的预测模型对钢企能耗进行预测。最后与BP神经网络以及最小二乘法进行比较。仿真结果表明该混合算法增强神经网络的泛化能力,具有相对误差小,预测精度较高,能更好地跟踪未来数据的优点。

Research of steel production consumption forecast based on simulated annealingPSO-BP algorithm

Against the existing problems in energy modeling and forecasting methods of BP neural networks, this particle propose a particle swarm optimization(PSO) based on simulated annealing(SA) to optimize BP neural network, combining with advantages of the simulated annealing algorithm and particle swarm algorithm ,which apply to optimize the weights threshold of BP neural network. Then the optimized network is trained ,and establish a model for predicting Guangzhou iron and steel group energy consumption. Comparing with BP neural network and least square method, the results of the simulation show that the hybrid algorithm can enhance the generalization ability of neural networks and has the advantage of smaller error, higher precision and better tracking.