Modeling of wind speed spatio-temporal series by multivariate-GARCH and copula/GARCH models

Abstract

This paper discusses the application of five t-GARCH models to the problem of accurately modeling three univariate but mutually dependent wind speed series taken from three US metering sites distant few kilometers from each other. Besides a benchmark model consisting of three independent univariate t-GARCH models, a t-CCC, a t-DCC, a t-copula/t-CCC and a t-copula/t-DCC model will be estimated, studied in their unconditional (i.e. static) and conditional (i.e. fully dynamic) statistical features, and compared to each other and to some statistical features of the original series. In order to highlight the usefulness of choosing volatility-oriented modeling such as\ud multivariate GARCH modeling for wind speed series, an Energy Finance application of capital budgeting under risk, i.e. energy portfolio selection, will be discussed and applied to the five\ud modeling schemes

Scenario generation and probabilistic forecasting analysis of spatio-temporal wind speed series with multivariate autoregressive volatility models

“…This approach comes mainly from financial econometrics literature. Such a multivariate dynamic modeling frame, preliminarily discussed in [13] in relation to wind speed time series modeling, seems not having ever been considered in the wind literature, nor having ever been investigated in that context in any theoretical and numerical depth. This quasi-linear approach is partly discriminative and partly probabilistic, because it dynamically models some parameters of its conditional distributions, i.e.…”

“…Models like GARCH-copula models, which can be considered even closer to a fully probabilistic approach, could also be tried, but from the preliminary analysis carried on in Ref. [13] it seems that they don't add very much to what more simple and standard mGARCH models like a DCC can already do. Models like switching mGARCH models [22] could on the contrary be very interesting in this context, but they are much more difficult to estimate.…”