Correction of Outliers in Temperature Time Series Based on Sliding Window Prediction in Meteorological Sensor Network

Li Ma, Xiaodu Gu, Baowei Wang
2017 Information  
In order to detect outliers in temperature time series data for improving data quality and decision-making quality related to design and operation, we proposed an algorithm based on sliding window prediction. Firstly, the time series are segmented based on the sliding window. Then, the prediction model is established based on the history data to predict the future value. If the difference between a predicted value and a measured value is larger than the preset threshold value, the sequence
more » ... will be judged to be an outlier and then corrected. In this paper, the sliding window and parameter settings of the algorithm are discussed and the algorithm is verified on actual data. This method does not need to pre classify the abnormal points and perform fast, and can handle large scale data. The experimental results show that the proposed algorithm can not only effectively detect outliers in the time series of meteorological data but also improves the correction efficiency notoriously. frequency is up to the minute level. Towards the end of 2012, the number of China's AWSs has reached 46,000. However, the average distance of AWS is only about 20 km. For densely populated China, this coverage is far from enough. For modern weather forecasts, it is necessary to have enough accurate weather information for improving decision-making quality related to design and operation. The unique characteristics of wireless sensor networks (WSN) [5, 6] have been favored by the scientific community and the military. It has been widely used in many fields, such as battlefield monitoring [7], environmental protection [8] , industrial control [9] and so on. The meteorological sensor network is defined as a network composed of meteorological sensor nodes, sink nodes, wireless communication facilities, and so on [10] . It can monitor and collect many kinds of weather information, such as temperature, humidity, air pressure, and wind speed. The obtained information will be processed and transmitted by a wireless multi hop mode, and finally sent to the control center by the sink nodes. Loading cheap weather sensors on nodes in meteorological sensor networks can greatly
doi:10.3390/info8020060 fatcat:mp7rtfg3enghfjdjacudntqkfm