基于GEO-GRU的电梯滑移量预测方法

A prediction method of elevator slip based on GEO-GRU

电梯曳引性能是影响电梯运行安全的重要指标,而滑移量是其重要的评判标准。为实现高精度的电梯滑移量预测,文中提出了基于金鹰算法优化的门控循环单元神经网络(GEO-GRU)模型,GRU通过门控单元实现数据信息选择,利用神经元传递时序特征,实现时序预测;采用金鹰优化算法(Golden Eagle Optimizer,GEO)对GRU的初始参数进行优化,以提升模型可靠性和泛化性。实验结果表明,相比LSTM、ESN等经典预测方法,GEO-GRU方法提高了预测精度,缩小了预测误差,对于电梯滑移量的预测具有更高的可行性和有效性。

The traction performance of elevator is an important index that affects the safe operation of elevator,and the slip is its important evaluation standard.In order to achieve high-precision elevator slip prediction,a gated cyclic unit neural network(GEO-GRU)model optimized by Golden Eagle Optimizer is proposed.GRU selects data information through gating unit,and uses neurons to transmit time series characteristics to make time series prediction.Golden Eagle Optimizer(GEO)is used to optimize the initial parameters of GRU to improve the reliability and generalization of the model.The experimental results show that compared with LSTM,ESN and other classical forecasting methods,GEO-GRU method improves the forecasting accuracy,reduces the forecasting error,and has higher feasibility and effectiveness in forecasting elevator slip.