Data Collection and Analyzing for IoT

  • Data Collection for Distributed Generator

As distributed generators become more and more popular, data collection for these generators becomes more important for father researches about data analysis. By considering the distributed generators as a wireless sensor network(WSN), it is easier to process energy efficient data collection strategies  such as clustering algorithms, data compress algorithms, mobile sink collection mechanism and so on. Clustering algorithms would divice the WSN into several clusters, each cluster has a cluster head(CH) which is in charge of this cluster’s data collection. By doing this, generator don’t have to transmit its data by multi-hops or long distance transmission. Date compress is usually processed in the CH because CH is a sub-data center in WSN. Mobile sinks are used to collect data from CHs. This would save energy compared with transmitting data directly to data center.


  • AI Based Data Analyzing for Energy Switch

The forecasting of power generation and user behavior analysis are of great significance for the planning of dispatching and optimal operation of energy switch. Machine learning, technological advancement in the field of artificial intelligence provides an effective way to refine the influencing factors of power generation prediction and user behavior analysis. By processing machine learning algotithms on data collected from generators and loads, we can extract many features of these data which are important for making decisions for power dispather.


  • Effective Energy Management for Energy Switch

The optimization problem of energy efficiency for data centers has been paid widespread attention. We investigate this problem in a new idea under the background of energy Internet, where subscribers are equipped with storage and smart energy management devices, especially for industry subscribers. In addition, there are large scale of clean energy generation and electricity-sale companies, which means that industry subscribers can purchase electricity from multi-source suppliers to cut down their energy cost and improve their energy efficiency based on real-time price, pollution index, etc. It is assumed that data centers always attempt to choose cheaper and cleaner energy in each hour and buy more electricity in valley hours with lower price compared to peak hours to reduce the energy cost. Thus both of pollution index function and real-time price are adopted to formulate the multi-source energy-purchasing cost. And both of the operation cost and potential cost are adopted to model the charging and discharging cost of storage devices, or storage cost for short. Based on this, the energy cost model with storage is formulated and is compared with the one without storage. We give the related algorithm to solve these problems and give the analysis of performance.


  • Group Paging Method for Cellular IoT

To handle the massive machine type communications (mMTC) and alleviate congestion of RAN, the group paging (GP) scheme was proposed. However, its performance quickly decreases in the face of massive simultaneous channel accesses. Hence, we proposed a two phase cluster-based group paging (CBGP) scheme.

  • Firstly, owing to the advantages of low cost, high access capacity and handy deployment, IEEE 802.11ah is introduced to increase the capability of coping with massive access attempts.
  • And the separation of inner-cluster data collection and header-based data transmission phases greatly alleviates access congestion of cellular networks, reducing the access delay and increasing the successful access probability for mMTC devices.
  • Besides, mathematical models of the CBGP scheme are derived in terms of the successful access probability and average access delay.
  • Finally, effects from different numbers of clusters on the performance of the CBGP scheme are investigated and the optimal number of clusters is also derived, adaptive to different access scales.