Event Title

Data Center Load Forecast using Hidden Markov Models

Location

University Student Union: Volstorff A

Start Date

12-2-2018 12:00 PM

Description

The energy cost of data centers tantamount to their overall operational cost. A possible solution to this immense cost could be proper scheduling of the power resources. This can be achieved by forecasting the data center loads. However, highly variable nature of the data center loads makes it challenging to use the traditional methods of load forecasting. In this paper, a stochastic method based on Hidden Markov process is developed to model the data center load and is used for a day-ahead forecasting. This method is out-standing because of its flexibility in addressing the variable nature of the data center load. The utility of the model is illustrated using a dataset from National Renewable Energy Laboratory - Research Support Facility (NREL - RSF). Two models created based on the proposed method yielded Mean Absolute Percentage Errors (MAPE) of 1.49% and 3.89%.

This document is currently not available here.

Share

COinS
 
Feb 12th, 12:00 PM

Data Center Load Forecast using Hidden Markov Models

University Student Union: Volstorff A

The energy cost of data centers tantamount to their overall operational cost. A possible solution to this immense cost could be proper scheduling of the power resources. This can be achieved by forecasting the data center loads. However, highly variable nature of the data center loads makes it challenging to use the traditional methods of load forecasting. In this paper, a stochastic method based on Hidden Markov process is developed to model the data center load and is used for a day-ahead forecasting. This method is out-standing because of its flexibility in addressing the variable nature of the data center load. The utility of the model is illustrated using a dataset from National Renewable Energy Laboratory - Research Support Facility (NREL - RSF). Two models created based on the proposed method yielded Mean Absolute Percentage Errors (MAPE) of 1.49% and 3.89%.