Demand curve prediction via Bayesian probability assignment over a functional space
Michael G. Traverso, Ali E. Abbas · 2009
One of the important aspects of energy modeling is the process of demand curve prediction. Existing demand curve prediction methods generally rely on statistical curve fittings which assume a certain functional form such as constant price elasticity. There are a number of disadvantages to this approach. For one, this method makes certain assumptions about the functional form of the price-demand curve that may not be exhibited in practice. In addition, since curve fits rely on only a single function, and not a distribution of functions, they do not capture the uncertainty about price-demand curves. In this work, demand curve prediction is instead treated by assigning a probability measure to the space of all functions that meet the global regularity (non-decreasing conditions). Using this method, a numerical example of Bayesian demand curve prediction is presented.