Power Consumption Model of a Microgrid Based on Utility-Providing History Data
This paper develops an energy consumption model for a microgrid based on utility-providing records and evaluates its accuracy. Built upon a data processing framework totally consisting of free open software, the well-known auto-regressive integrated moving average method is exploited for time series tracking. The series of quarter-hourly records during 60 days are fed to the modeling process, and the consumption of thenext 1/4, 1, 2, and 3 hours is predicted. The average prediction error of 15-minute-ahead forecast remains below 3.4 kwh, while that of the 3-hour-ahead forecast is extended to 12.37 kwh. Next, day-by-day measurement shows that the 15-minute-aheadprediction has about just 1.5 % for the whole period in spite ofa significant change between weekdays and weekends. This bigdata-based prediction can allow us to build a new intelligent energy application such as battery controllers and electricity trade coordinators.
Keywords: Microgrid, Power Consumption History, ARIMA, Prediction Accuracy, Statistics Package.