【作者】 张涛；

【导师】 钟俊； 崔开源；

【作者基本信息】 四川大学 ， 电气工程（专业学位）， 2021， 硕士

【摘要】 智能电网的发展、自动抄表技术的进步对于电力营销系统具有极其重大的促进作用。而智能电网与自动抄表技术要解决的两个核心问题在于,如何使系统运行更快更准确,以及如何提供更智能化的服务。只有更快更准确地传输、处理、相应复杂的电力网络中产生的数据,才能够提高系统的实时性,提供更好更便捷的服务。传统电网中采用的窄带电力线载波通信方式、光纤通信方式等通信手段传输效率低,信道架设成本高。而基于正交频分复用调制技术(Orthogonal Frequency Division Multiplexing,OFDM)的宽带高速电力线载波通信(High-speed Power Line Carrier Communication,HPLC)兼具了高传输速率、高准确率、高稳定性等优点,而且利用电力网络中随处可见的电力线作为信道,可以节省成本,是智能电网中理想的通信方式。同时,因为电网规模庞大,网络边缘产生、收集数据的设备众多,各种设备之间异构性很高,而引入边缘计算可以在网络边缘对这些数据进行预处理、建模分析、发出指令、上传云端,不仅能够提高数据的利用率,将处理数据的任务从网络中心转移到网络边缘也减少了云端的计算量以及出入云端的通信量,避免了在云端出现故障时整个网络瘫痪而造成的损失。基于此,本论文提出了一种利用HPLC和边缘计算模块来收集、传输、处理、响应用电数据的用电信息采集系统。主要工作如下:(1)介绍了电力线载波通信的基本原理以及信道特点,分析了其噪声产生的原因以及相应的解决措施。在PLC的基础上利用OFDM调制来解决PLC传输稳定性的问题,并给出了利用OFDM调制的HPLC通信系统的物理层结构,对物理层中信号处理的模块进行了介绍。本论文所采用的HPLC通信系统相比传统PLC通信,通信带宽更宽,码间串扰、信号衰减、噪声干扰更小,传输更为稳定,可以满足现代智能电网的通信需求。最后,提出了一个基于电力线载波特征值的台区识别算法。(2)为了使营销系统能够更好地发挥作用,提出了在边缘计算模块中利用长短时记忆(Long Short Term Memory,LSTM)对收集到的用电数据、气候数据进行建模分析,对短期内的电网负荷进行预测。本论文针对用电数据的特点建立了模型并进行了实验验证,证明了所建立模型对比起其他方法在负荷预测上具有优越性。针对电网中经常存在的窃电等异常现象,提出了基于LSTM算法的异常检测模型,对用电数据集进行了标注建模并进行了测试。经过测试,该模型异常检测成功率高达90%以上,远超人工检测的成功率,大大节省了人工成本,提高了电网系统工作效率。(3)分析了智能电网系统中数据处理的需求,提出了基于HPLC通信的边缘计算模块的设计及集中器设计。更多还原

【Abstract】 The development of smart grid and the advancement of automatic meter reading technology play an extremely significant role in promoting the power marketing system.The two core problems to be solved by smart grid and automatic meter reading technology are how to make the system run faster and more accurately,and how to provide smarter services.Only faster and more accurate transmission and processing of the data from complex power network,can the real-time performance of the system be improved with better and more convenient services.Communication methods such as narrowband power line carrier communication and optical fiber communication used in traditional power grids have low transmission efficiency and high channel construction costs.Based on Orthogonal Frequency Division Multiplexing(OFDM)technology,the broadband High-speed Power Line Carrier Communication(HPLC)combines the advantages of high transmission rate,high accuracy,and high stability,and uses the power lines everywhere in the power network as a channel to save costs It is an ideal communication method in the smart grid.At the same time,because of the large scale of the power grid,there are many devices that generate and collect data at the edge of the network,and the heterogeneity between various devices is high.The use of edge computing can preprocess,model and analyze these data,as well as sending instruction and upload to the cloud at the edge of the network.This can not only improve the utilization of data,but also transfer the task of processing data from the network center to the edge of the network.It also reduces the amount of cloud computing and the amount of communication to and from the cloud,avoiding the loss caused by the whole network paralysis when the cloud fails.Based on this,this paper proposes an electricity consumption information collection system that uses HPLC and edge computing modules to collect,transmit,process,and respond to electricity consumption data.The main tasks are as follows:(1)The basic principles and channel characteristics of power line carrier communication are introduced,and the causes of noise and corresponding solutions are analyzed.On the basis of PLC,OFDM modulation is used to solve the problem of PLC transmission stability,and the physical layer structure of the HPLC communication system using OFDM modulation is given,and the signal processing module in the physical layer is introduced.Compared with traditional PLC communication,the HPLC communication system used in this paper has wider communication bandwidth,smaller inter-code crosstalk,signal attenuation,and less noise interference,and has more stable transmission,which can meet the communication needs of modern smart grids.Finally,a station recognition algorithm based on power line carrier eigenvalues is proposed.(2)In order to make the marketing system work better,it is proposed to use Long Short Term Memory(LSTM)in the edge computing module to model and analyze the collected electricity consumption data and climate data,and to analyze the short-term predict the load within the grid.In this paper,a model is established based on the characteristics of electricity consumption data and verified by experiments,which proves that the established model is superior to other methods in load forecasting.Aiming at the abnormal phenomena such as electricity theft that often exist in the power grid,an anomaly detection model based on the LSTM algorithm is proposed,and the electricity consumption data set is labeled and modeled and tested.After testing,the model has an abnormal detection success rate of more than 90%,far exceeding the success rate of manual detection,greatly saving labor costs and improving the efficiency of the power grid system.(3)The data processing requirements in the smart grid system are analyzed,and the design of edge computing module and concentrator based on HPLC communication are proposed.更多还原