Probabilistic power forecast of renewable distributed generation for provision of control reserve using vine copulas

Abstract

This work focuses on calculating the amount of control reserve, which can be provided by a pool of renewable power plants on the next day. The power forecast of wind and solar power plants depends on the weather forecast, which always contains errors. A merger of individual plants at different locations is advantageous in order to reduce the overall forecast error. Still, the amount of control reserve needs to be determined with a high level of reliability. For the calculation, a probabilistic approach based on historical and current weather data is chosen. In order to model the spatial dependencies of the forecast errors between individual plants, R‐vine copulas are used. In the copula theory, R‐vine copulas provide high accuracy in modelling the dependency of stochastic variables. The methodology is validated and compared to three alternative approaches with a use case of 32 wind and solar plants. The calculated amount of control reserve provided and the achieved reliability proves to be superior to alternative approaches. Additionally, the required reliability level is varied to investigate the impact on the amount of control reserve, which can be offered with the pool.

IEA Wind Recommended Practice for the Implementation of Renewable Energy Forecasting Solutions

“…To be able to benefit from integration of information from geographically distributed time series or from a grid of NWP the method needs to add a second dimension. Another approach is to use copula theory, which offers a flexible approach to the probabilistic power forecast taking the spatial correlation of the forecast errors into account [13,14].…”

“…In this approach scenarios are produced by a statistical scenario generation method from the probability distributions produced by statistical models based on the copula theory [13,14]. We define them as scenarios, as the further processing of the approach contains x independent results in contrast to the statistical method, producing a PDF function.…”

Data‐driven look‐ahead voltage stability assessment of power system with correlated variables

“…In addition, PDF forecasting methods provide more information about future uncertainties of loads and generations. KDE and its modified versions are well utilized for load forecasting [29], and generating wind [30] and PV [31]. However, KDE may suffer from the boundary effect for bounded variables.…”