Influential factors for accurate load prediction in a Demand Response context

Accurate prediction of a buildings electricity load is crucial to respond to Demand Response events with an assessable load change. However, previous work on load prediction lacks to consider a wider set of possible data sources. In this paper we study different data scenarios to map the influence of the different data parameters. We also look at the temporal aspect of predicting by looking at the predicted seasons. By predicting with a MultiLayer Perceptron, which is a universal approximator, it is possible to focus solely on the influence of the parameters instead of the prediction algorithm itself. Finally, multiple prediction algorithms are compared. The influential factor analysis is based on data from an entire year from a office building in Denmark. The results show that weather data is the most crucial data parameter. A slight improvement from load data was however seen using only occupancy data. Next, the time of day that is being predicted greatly influence the prediction which is related to the weather pattern. By presenting these results we hope to improve the modeling of building loads and algorithms for Demand Response planning.

[1]  Mikkel Baun Kjærgaard,et al.  Improving occupancy presence prediction via multi-label classification , 2016, 2016 IEEE International Conference on Pervasive Computing and Communication Workshops (PerCom Workshops).

[2]  Michael Y. Hu,et al.  Forecasting with artificial neural networks: The state of the art , 1997 .

[3]  Herbert Jaeger,et al.  Reservoir computing approaches to recurrent neural network training , 2009, Comput. Sci. Rev..

[4]  Guy R. Newsham,et al.  Building-level occupancy data to improve ARIMA-based electricity use forecasts , 2010, BuildSys '10.

[5]  Heaton T. Jeff,et al.  Introduction to Neural Networks with Java , 2005 .

[6]  Dianhui Wang,et al.  Extreme learning machines: a survey , 2011, Int. J. Mach. Learn. Cybern..

[7]  Bernhard Schölkopf,et al.  A tutorial on support vector regression , 2004, Stat. Comput..

[8]  Donald F. Specht,et al.  A general regression neural network , 1991, IEEE Trans. Neural Networks.

[9]  Martin A. Riedmiller,et al.  A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[10]  Bernard Widrow,et al.  30 years of adaptive neural networks: perceptron, Madaline, and backpropagation , 1990, Proc. IEEE.

[11]  Eric Wai Ming Lee,et al.  An intelligent approach to assessing the effect of building occupancy on building cooling load predi , 2011 .

[12]  Nikolaos Kourentzes,et al.  Neural network ensemble operators for time series forecasting , 2014, Expert Syst. Appl..

[13]  Tin-Tai Chow,et al.  The use of occupancy space electrical power demand in building cooling load prediction , 2012 .