CONVERGENT CROSS MAPPING FOR DEMAND FORECASTING

The availability of inexspensive electricity in a constant and reliable fashion is critical to economic development and efficient resource consumption. To this end, accurate short term load forecasting (STLF) on an electrical grid enable the minimization of dispatch and running costs on the scale of seconds to a week. Models and approaches employed in STLF include multi-linear regression, Box-Jenkins Analysis, fuzzy systems, non-linear state space reconstruction (SSR), and various hybrid models. The domain of this project lies in pure and potentially hybrid non-linear state space methods where SSR have already been explored (5,7). Unexplored in this domain is a method called convergent cross mapping (CCM). The thrust of CCM is a potentially novel approach to making inferences regarding the causal drivers of a time series variable. The primary goal of this project is: use CCM to identify predictors of power grid demand and determine whether or not such predictors improve upon current demand forecast methods.

Faculty Supervisor:

Simon Bonner

Student:

Partner:

University of California, San Diego

Discipline:

Mathematics

Sector:

Education

University:

The University of Western Ontario

Program: