Path Analysis And Energy Consumption Prediction Based On Hadoop

With the rapid development of urban road traffic, the energy consumption and traffic pollution caused by traffic also mount sharply, becoming main obstacles restricting the development of the city and enormous challenges to the traffic management. To solve these problems, great efforts have been made by our country to develop the intelligent transportation system, which applies a new generation of information technology, such as the mobile Internet, Internet of things and cloud computing, deeply into the traffic field. Meanwhile, massive data has been gained in the area of intelligent transportation, and the data research and development based on the new generation of information technology have brought people an era with both opportunities and challenges.As previous data processing tools and techniques have been unable to solve current problems, there is an urgent demand to seek for new technology to handle the big data. The Hadoop is an open source, distributed cloud computing framework, owning a great advantage in terms of massive data storage and processing. In this paper, we use Hadoop to process and analyze massive floating car data and energy consumption data, and then study the relationship between the average speed of road section and the average energy consumption per hundred kilometers. Therefore, rough forecast of the average energy consumption per hundred kilometers can be made and knowledge of the relation between air quality and energy consumption could be gained at the same time.In this dissertation, preprocessing of the massive floating car data and the energy consumption data was carried out by Shell and Hadoop to merge and remove the duplicate data items, which ensured high data availability. Using Hive to deal with the road network data and calculate the direction angle of roads, and then the map data file was established. On this basis, we moved the map matching method based on matching degree to the Hadoop platform, and the latter could implement the matching of large-scale floating car data and the energy consumption data based on its distributed cache technology.On the basis of accurate map matching of floating car data and energy consumption data, the link average velocity is calculated using improved velocity time integral model while link vehicle average fuel consumption per hundred kilometers is calculated using the cumulative method. At the same time, the two algorithms are transplanted to Hadoop platform to calculate the large-scale floating car data as well as the link average speed and energy consumption per hundred kilometers data. Period of time is separated into unit of month to calculate the average speed of certain links and the average fuel consumption per hundred kilometers.The prediction of link average energy consumption per hundred kilometers was conducted using BP neural network and the result was analyzed and validated by fitting of real data, which proved good effectiveness and validity. The fitting curves were then analyzed and validated by reserved data. Finally, the total fuel consumption data calculated by Hadoop was gained and researched in terms of its relationship with AQI and PM2.5 in the haze, the results of which were analyzed as well.