基于LSTM神经网络机车调度剩余时间预测  

Prediction of Remaining Time of Locomotive Scheduling Based on LSTM Neural Network

作　　者：方欢 张守政[1] FANG Huan;ZHANG Shouzheng(College of Mathematics and Big Data,Anhui University of Science and Technology,Huainan Anhui 232001,China;Anhui Province Engineering Laboratory for Big Data Analysis and Early Warning Technology of Coal Mine Safety,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学数学与大数据学院,安徽淮南232001 [2]安徽理工大学安徽省煤矿安全大数据分析与预警技术工程实验室,安徽淮南232001

出　　处：《安徽理工大学学报（自然科学版）》2022年第5期63-70,共8页Journal of Anhui University of Science and Technology:Natural Science

基　　金：国家自然科学基金资助项目(61902002)。

摘　　要：矿山机车调度的数字孪生系统构建是智慧矿山建设的一项重要研究内容,为实现矿山机车调度数字孪生系统的实时预测分析,将预测型监控管理的剩余时间预测作为研究目标。从井下机车调度系统的数字孪生系统的背景出发,以矿山机车的智能传感技术实时采集的机车运行过程数据为基础,通过对事件日志中不同类型的机车轨迹进行划分,构建剩余时间向量,并使用对时序数据有较高准确性的LSTM神经网络进行预测,实现机车调度剩余时间的精准预测。同时,根据划分的机车轨迹创建滑动窗口矩阵,在机车运行过程中不断迭代更新,使用已训练的LSTM神经网络模型对滑动窗口矩阵的剩余时间向量进行预测,实现不同类型机车调度剩余时间预测的实时更新。通过实验证明了方法的有效性。提出的方法可以有效地解决不同机车类型在不同活动下的机车调度剩余时间实时预测问题,为矿山数字孪生下的机车大数据分析与决策提供数据支持。The construction of the digital twin system of mine locomotive dispatching is an important research content in the construction of smart mines.In order to realize the real-time prediction and analysis of the digital twinsystem of mine locomotive dispatching,the remaining time prediction of predictive monitoring management was taken as the research goal.Based on the process data of the locomotive operation in real time collected with the intelligent sensing technology of the mining locomotive in the background of the digital twin system of the underground locomotive dispatching system,the remaining time vector was constructed by dividing the different types of locomotive traces in the event log.The LSTM neural network with high accuracy of time series data was used to make predictions to achieve accurate prediction of the remaining time of locomotive scheduling.At the same time,a sliding window matrix was created based on the divided locomotive trace,which was continuously updated iteratively during the operation of the locomotive.The remaining time vector of the sliding window matrix was predicted by the trained LSTM neural network model,which realizedthe real-time updates of the remaining time forecasts of different types of locomotive dispatching and provided data support for locomotive big data analysis and decision-making under the digital twin of the mine.The proposed method is able to solve effectively the problem of real-time prediction of locomotive scheduling remaining time under different activities for different locomotive types.

关 键 词：数字孪生系统 机车调度 LSTM神经网络 剩余时间向量 剩余时间预测 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]