Smart Grid Oriented Data Visualization Computation On Energy Theft Detection

This paper tries to solve problem of illegal electricity detection within smart power grid, Illegal electricity usage means power grid end consumers steal energy via illegal method. Currently, people solve this problem by either tempering power device physically or classification method with machine learning. But with a power grid which has complex structure and massive data, traditional methods can’t handle it well along with vari-eties of energy theft methods:firstly they can’t cover the whole grid; sec-ondly they response slowly, in some case, the duration could be a couple of months; and finally the accuracy ratio of traditional detection mechanism is less than 60%, which is not good enough.Aiming at the problem described above, this paper tries to visualize power grid data, and get power grid data analyzing experts into the process of final verdict, experts read pre-alert information of suspicious districts or households, determine whether this particular behavior is illegal or not; meanwhile we design new detection algorithm, and simplify the topology of power grid which is an input to new algorithm, thus to generate energy theft suspects list as quickly as possible, this could be pre-alert for post verdict. Based on this design theory, this paper proceeds primary research work as following:a) Design and implement brand-new distributed de-tection algorithm based on linear equation system solving methods, build the whole distributed pre-alert system with fast response speed; b) Build electricity data storage platform to handle massive data storage and access control within smart power grid with the help of open source distributed storage infrastructure; c) Build the whole visualized global power grid data monitoring and analyzing platform which allows experts to get access to from anywhere.Finally, we exploit Mosaik simulation framework to simulate some real scenarios of illegal energy usage, thus to test the performance of whole detection system. Experiment data shows that:a) This new detection sys-tem can cover 100 percent of power grid naturally; b) Our detection system can give a feedback within half a day even when it comes to large power grid consists of more than 10,000 households; c) The accuracy ratio of de-tection can still be larger than 70%in the magnitude of 10,000 households. This distributed illegal energy usage detection and data visualization plat-form proposed by this paper has showed a significant improvement in grid covering ratio, response time and accuracy ratio.