Design And Implementation Of Power User Behavior Analysis Platform For Power Grid Big Data

With the development of society,electricity plays an increasingly important role in people’s life,where both living and entertainment are inseparable from the power system.In recent years,the smart grid has been widely popularized and the power industry has entered the era of big data since the Internet of Things developed rapidly.Power companies own a large amount of user load data so how to find out its value will become the key to change future energy market.Power load analysis mainly analyzes the data by means of machine learning,trying to discover the operation laws,which mainly includes power consumption prediction,user clustering,user portrait analysis and so on.It can be used to provide data support for specific services such as step tariff design,abnormal user detection and load balancing implementation.This article aims to design and implement a power user behavior analysis platform for power grid big data.It is designed to provide behavior analysis services based on actual business scenarios,which can select algorithms based on scenario characteristics to analyze power data and provide data visualization.Compared with existing analysis platforms,the system is connected with business scenarios more closely and the data has a more interpretable physical meaning,which is more convenient for business personnel to understand and process.Besides,this article proposes several improvements to existing data analysis algorithms,which solves their disadvantages and makes them more suitable for time series data analysis.Finally,the system provides good scalability which can support access to new algorithms and scenarios in the future.In order to realize this behavior analysis platform,this article first investigates the relevant technologies required for system implementation and analyzes the feasibility of system design.Secondly,by analyzing the requirements of the system,the typical application scenarios of the system are clarified and two scenarios including user similarity analysis and abnormal user detection are selected for the system.Then,based on specific scenario requirements,two key issues that need resolving are analyzed and settled:On the one hand,in order to measure the similarity of time series data,this paper proposes a user similarity analysis method based on limited DTW algorithm,which introduces search constraints and translation constraint to improve the computing efficiency while also adapting to the characteristics of power data.On the other hand,so as to detect potential abnormal users,this paper proposes an abnormal user detection method based on multiple time scales,which combines short-term power consumption patterns with long-term user behaviors to comprehensively measure whether users are abnormal.After that,the overall design of entire behavior analysis platform is carried out while the system is divided into different modules and key modules are designed in detail.Finally,the platform is built with its all functions tested thoroughly.At the end of article,the overall work is summarized and the remaining shortcomings of behavior analysis platform and future improvement directions are pointed out.