没有合适的资源?快使用搜索试试~ 我知道了~
第十五届中国研究生电赛_继续奔跑的小白1
需积分: 0 5 下载量 123 浏览量
2022-08-03
19:58:32
上传
评论
收藏 4.8MB PDF 举报
温馨提示
试读
42页
摘要在无线通信网络的规划与建设中,高效精确的无线链路损耗预测对于基站网络优化有着重要的指导意义。无线信道的电波传输特性与信道环境密切相 关。复杂的电磁通信环境下
资源详情
资源评论
资源推荐
第十五届中国研究生电子设计竞赛
技术论文
中文题目:数据驱动的机器学习算法为基站链路损耗
预测与站址规划赋能
英文题目:Data-driven machine learning
algorithms empower base station link
loss prediction and site planning
参赛单位:南京航空航天大学
队伍名称:继续奔跑的小白
指导老师:李海林
参赛队员:杨凌辉、杨昌林、张嘉纹
完成时间:2020-7-20
摘 要
在无线通信网络的规划与建设中,高效精确的无线链路损耗预测对于基站网
络优化有着重要的指导意义。无线信道的电波传输特性与信道环境密切相 关。
复杂的电磁通信环境下,会产生诸如反射、散射、绕射等电波传播方式,此时的
链路损耗与自由空间损耗模型大有不同,需要加入特定的场景修正因子。精准的
无线电传输损耗模型的建立,可以对基站目标通信覆盖范围内的电磁传输情况进
行相对准确的预测,为后续通信业务指标例如:小区基站覆盖范围、小区间网络
干扰、通信传输速率以及通信链路容量等指标的有效估算提供理论依据。传统电
波传输损耗模型主要有三种,经验模型、理论模型和改进经验模型。经验模型的
设计是根据大量场景实测数据进行公式参数拟合,典型模型有 Cost231-Hata、
Okumura 等。理论模型主要依据电磁传播理论、考虑电磁波在空间中的反射、折
射和散射,借助几何光学理论、几何绕射理论和一致绕射理论等,进行理论建模,
代表性的是 Volcano 模型。改进经验模型是依据特定电波传输场景下的实测数据,
设定特定的场景修正因子,从而获得更为细化的分类场景传输模型,典型的有
Standard Propagation Model, SPM。近年来随着通信设备数量的增多,以及大数据
云计算等存储能力和算力的应用和普及,产生了大量的基站信号测量数据,本文
基于丰富的实测数据,提出了基于随机森林算法 RF (Random Forests, RF)、CNN
(Convolutional Neural Network, CNN) 和 DNN (Deep Neural Network, DNN) 网络
的链路损耗预测模型,与传统模型预测精度进行对比分析。并在此基础上,创新
性地引入基于加权聚类算法的 KW-means 模型对基站的最优部署进行优化,结果
表明,该算法可以有效减少基站弱信号覆盖比例,从而更好地服务基站用户。
关键词:链路预测,聚类算法,站址规划,RF,CNN,DNN
Abstract
In the planning and construction of wireless communication networks, efficient and
accurate wireless link loss prediction has important guiding significance for the
optimization of base station networks. The radio wave transmission characteristics of
wireless channels are closely related to the channel environment. In a complex
electromagnetic communication environment, there will be radio wave propagation
methods such as reflection, scattering, and diffraction. At this time, the link loss is very
different from the free space loss model, and a specific scene correction factor needs to
be added. The establishment of an accurate radio transmission loss model can make
a relatively accurate prediction of the electromagnetic transmission within the target
communication coverage of the base station, for subsequent communication business
indicators such as cell base station coverage, inter-cell network interference,
communication transmission rate and communication chain The effective estimation
of indicators such as road capacity provides a theoretical basis. There are three main
models of traditional radio wave transmission loss, empirical model, theoretical model
and improved empirical model. The design of the empirical model is based on a large
number of scene measurement data for formula parameter fitting. Typical models
include Cost231-Hata and Okumura. The theoretical model is mainly based on the
theory of electromagnetic propagation, considering the reflection, refraction and
scattering of electromagnetic waves in space. The theoretical modeling is carried out
with the help of geometric optics theory, geometric diffraction theory and uniform
diffraction theory. The representative is the Volcano model. The improved empirical
model is based on the measured data in a specific radio wave transmission scenario,
setting a specific scene correction factor, so as to obtain a more detailed classification
scene transmission model, typical are Standard Propagation Model, SPM. In recent
years, with the increase in the number of communication devices, and the application
and popularization of storage capacity and computing power such as big data cloud
computing, a large amount of base station signal measurement data has been generated.
Based on rich measured data, this paper proposes network link loss prediction models
based on random forest algorithm (Random Forests, RF), CNN (Convolutional Neural
Network, CNN) and DNN (Deep Neural Network, DNN) ,they are compared with the
prediction accuracy of traditional models. On this basis, the KW-means model based
on the weighted clustering algorithm is innovatively introduced to optimize the optimal
deployment of the base station. The results show that the algorithm can effectively
reduce the weak signal coverage ratio of the base station to better serve the base station
users.
Keywords: link prediction, clustering algorithm, site planning, RF, CNN, DNN
目 录
第 1 章 作品难点与创新...................................................................................................1
1.1 作品难点 .......................................................................................................................1
1.2 作品创新点...................................................................................................................2
第 2 章 方案论证与设计...................................................................................................3
2.1 基站链路损耗数据集 ................................................................................................3
2.2 特征设计 .......................................................................................................................6
第 3 章 原理分析.................................................................................................................14
3.1 传统链路损耗模型...................................................................................................14
3.2 随机森林 .....................................................................................................................16
3.2.1 随机森林回归模型.......................................................................................16
3.2.2 模型参数分析.................................................................................................17
3.3 CNN 网络 ......................................................................................................................17
3.4 DNN 网络 ......................................................................................................................18
3.5 加权 KW-Menas 算法.................................................................................................19
第 4 章 软件设计与流程.................................................................................................21
4.1 随机森林模型 ............................................................................................................21
4.2 CNN 模型 ......................................................................................................................21
4.3 DNN 模型 ......................................................................................................................22
第 5 章 系统测试与误差分析......................................................................................26
5.1 随机森林模型预测结果..........................................................................................26
5.1.1 预测数据统计................................................................................................26
5.1.2 分布密度统计................................................................................................27
5.2 CNN 模型预测结果 ....................................................................................................28
5.2.1 预测数据统计................................................................................................28
5.2.2 分布密度统计................................................................................................29
5.3 DNN 模型预测结果 ....................................................................................................29
5.3.1 预测数据统计................................................................................................30
5.3.2 分布密度统计................................................................................................31
第 6 章 客户端设计 ...........................................................................................................32
6.1 系统架构图.................................................................................................................32
6.2 数据可视化客户端...................................................................................................32
6.3 站址规划客户端........................................................................................................32
第 7 章 总结 ...........................................................................................................................36
参考文献 ...................................................................................................................................37
剩余41页未读,继续阅读
黄涵奕
- 粉丝: 72
- 资源: 328
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0