Probabilistic description and prediction of electric peak power demand

Abstract

It is widely recognized that one of the crucial point for designing and planning electrical power system is the load characterization. The problem is well known and analyzed for any power system - since peak demand may exceed the maximum generated power, resulting in power outages and load shedding - but it is particularly cumbersome for railway and light transportation systems. In these cases indeed the loads exhibit fast changes and a large degree of randomness, whose description requires a proper analysis by using stochastic processes. In particular, it is of interest to have information about the extreme value of the stochastic load process in time for properly designing the generation and distribution system, and the storage devices. In the paper a new efficient estimation algorithm for the frequency of peak load occurrences is proposed. The core of the procedure, which is easily extensible to other peak load parameters, is based upon the assumption that the peak power is a Poisson process. In the paper, after a proper probabilistic description, attention is focused on the estimation of the above frequency by means of a suitable Bayesian estimation technique. Finally, the summary of a large set of numerical simulations is presented, which show the high efficiency of such estimation methodology.