Energy and Utilities Virtual Conference

Proceedings

A Predictive Model of Building Power Usage Through PI System and MATLAB Integration

Ahmad Fattahi and Michael Christopher, OSIsoft, LLC

Read abstract

In this presentation, we focus on showing the power of data. Using PI System as the enterprise data infrastructure, we collect samples of some physical quantities related to our office building, such as power consumption and outside temperature, on a regular basis. Over time, the data set becomes large. Leveraging the computational power of MATLAB, we integrate the two platforms, build a predictive model with MATLAB, funnel the data set, and train our model. In real time, we generate predictions of power consumption that incorporate data from a weather forecast web service. We also configure automatic notifications when power consumption deviates significantly from the "expected" level.