基于双池化的多层感知机光功率预测方法
杨菁1张欢1张烨宇1周驰楠2张晓龙 2
Optical Power Prediction Method Based on Double Pooling Multilayer Perception
YANG Jing1ZHANG Huan1ZHANG Yeyu1ZHOU Chinan2ZHANG Xiaolong2
-
作者信息:1.成都地铁运营有限公司， 610066， 成都
2.广西交控智维科技发展有限公司， 530025， 南宁
-
Affiliation:1.Chengdu Metro Operation Co., Ltd., 610066, Chengdu, China
2.Guangxi Jiaokong Zhiwei Technology Development Co., Ltd., 530025, Nanning, China
-
关键词:
-
Key words:
-
DOI:10.16037/j.1007－869x.2024.04.048
-
中图分类号/CLCN:U285
-
栏目/Col:产学研视窗
摘要:
［目的］针对地铁通信系统历史光功率数据缺失但历史相关数据充分的情况，提出基于双池化多层感知机模型的光功率预测方法。［方法］介绍了多层感知机模型和基于双池化多层感知机模型的结构和预测原理。基于成都地铁通信系统检测数据，通过试验对比分析了不同模型的预测结果，以及基于双池化多层感知机模型在不同变量下的预测结果。［结果及结论］基于双池化多层感知机模型具有较好的预测精度。在面对环境信息的多变性和时序信息的重复性问题，该模型能够分别使用最大池化(Maxpooling)和平均池化(Avgpooling)进行特征提取，有效利用数据信息对输入光功率进行预测；该模型充分考虑了地铁通信系统输入光功率受到的相关影响因素和数据的突变特点，能够有效解决光功率时常突变、难以预测的问题。
Abstracts:
［Objective］ In response to the situation where historical optical power data in metro communication systems is missing despite the sufficient historical related data, an optical power prediction method based on double pooling multilayer perceptron model is proposed. ［Method］ The structure and prediction principles of the multilayer perceptron model and the double pooling multilayer perceptron model are introduced. Based on detection data from the Chengdu Metro communication system, experimental comparative analysis is conducted on the prediction results of different models, as well as the prediction results of the double pooling multilayer perceptron model under different variables. ［Result & Conclusion］ The double pooling multilayer perceptron model demonstrates good prediction accuracy. In facing the variability of environmental information and the repetitiveness of time sequence information, the model can utilize maximized-pooling (Maxpooling) and average-pooling (Avgpooling) separately for feature extraction, effectively predicting input optical power with data information. The model fully considers the relevant influencing factors on the input optical power of metro communication system and the characteristics of data mutation, effectively addressing the problem of frequent and unpredictable optical power fluctuations.
过刊浏览
资料下载
友情链接