Presentation: Automated Pattern Recognition for PI Data Using MATLAB

In this talk we offer a solution to the problem of finding temporal patterns on PI tags. The user arbitrarily chooses a pattern of interest spanning a time interval from a PI tag. The user can then search for all the occurrences of similar patterns on any other PI tag over a bigger time interval, potentially years of archived data. The tightness of the search can be adjusted to fine-tune the results. The underlying analytics engine is MATLAB. We integrate PI System with MATLAB using a toolbox developed by one of OSIsoft System Integrators. The toolbox leverages PI SDK directly via COM, providing high performance data access that is intuitive to PI System and MATLAB users alike. We will discuss the topics of importing data from and returning results back to the PI System, as well as comparing the PI SDK approach to alternatives such as PI OLEDB.