摘要
电弧炉炼钢是一种复杂的工业生产过程,电极调节系统的性能是影响生产效益的重要因素。首先介绍了一种采用扩展DBD学习算法的电极神经网络预测控制方法,指出在实际应用中,这种控制方法存在着神经网络预测模型收敛速度慢的问题。针对此问题,提出了一种以改进GA和扩展DBD相结合的方法作为学习算法的预测控制方案,仿真结果表明,它可以较好地解决神经网络预测模型收敛速度慢的问题,并可以在一定程度上提高预测模型的输出精度。
Electric Arc Furnace steel-making is a complex industrial production process and the performance of its electrode adjusting system is a key factor which can effect the production benefit. The paper firstly introduces an electrode neural network predictive control method, which adopts the extended DBD algorithm as learning method, and points out that its convergent speed is slow in practice appliance. Thus, a new predictive control method, which combines the modified Genetic Algorithm with the extended algorithm, is presented consequently. The simulation result indicates that it can solve the slow convergent speed problem of neural network predictive model and improve the output precision of predictive model in a certain extent.
出处
《工业加热》
CAS
2006年第2期59-63,共5页
Industrial Heating
关键词
电弧炉
电极调节系统
预测控制
神经网络
扩展DBD算法
遗传算法
electric arc furnace
electrode adjusting system
predictive control
neural network
extended DBD algorithm
genetic algorithm