论文部分内容阅读
针对目前国内对铝电解槽运行状况诊断存在的难度大、效率低等问题,着眼于与实时工况相区别而反应电解槽电解能力和稳定运行的电解槽状态的研究,设计了一种以小波包算法提取槽状态信息和建立了用非线性Morlet小波基取代传统神经元非线性激励函数的紧致型小波神经网络的槽状态预测模型。利用小波变换的时域局部化性质和神经网络的自学习能力,对铝电解槽的槽状态进行分析预测,克服了传统神经网络收敛速度慢,容易陷入局部最优等缺点。通过Matlab对状态预测算法进行编程。结果显示,相比传统的神经网络预测模型,铝电解槽的槽状态预测更加准确。
In view of the current domestic problems in the diagnosis of the operating status of the aluminum reduction cell, which are difficult and inefficient, the research focused on the status of the electrolytic cell which is different from the real-time operation and reflects the electrolytic capacity of the electrolytic cell and the stable operation. The packet algorithm extracts the slot state information and builds a slot state prediction model of a compact wavelet neural network that replaces the traditional nonlinear neuron excitation function with a nonlinear Morlet wavelet basis. The time domain localization of wavelet transform and the self-learning ability of neural network are used to analyze and predict the slot state of aluminum reduction cell, which overcomes the shortcomings of traditional neural network, such as slow convergence speed and easy fall into local optimum. The state prediction algorithm is programmed by Matlab. The results show that the slot state prediction of aluminum reduction cell is more accurate than the traditional neural network prediction model.