Wind power scenario generation through state-space specifications for uncertainty analysis of wind power plants

• Published in: Applied energy 2016-01, Vol.162, p.21-30
• Main Authors: Díaz, Guzmán, Gómez-Aleixandre, Javier, Coto, José
• Format: Article
• Language: English
• Subjects: Multivariate stochastic processes; Simulation; State space; Wind power
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2015.10.052

[Display omitted] •State space representations for simulating wind power plant output are proposed.•The representation of wind speed in state space allows structural analysis.•The joint model incorporates the temporal and spatial dependence structure.•The models are easily integrable into a backward/forward sweep algorithm.•Results evidence the remarkable differences between joint and marginal models. This paper proposes the use of state space models to generate scenarios for the analysis of wind power plant (WPP) generation capabilities. The proposal is rooted on the advantages that state space models present for dealing with stochastic processes; mainly their structural definition and the use of Kalman filter to naturally tackle some involved operations. The specification proposed in this paper comprises a structured representation of individual Box–Jenkins models, with indications about further improvements that can be easily performed. These marginal models are combined to form a joint model in which the dependence structure is easily handled. Indications about the procedure to calibrate and check the model, as well as a validation of its statistical appropriateness, are provided. Application of the proposed state space models provides insight on the need to properly specify the structural dependence between wind speeds. In this paper the joint and marginal models are smoothly integrated into a backward–forward sweep algorithm to determine the performance indicators (voltages and powers) of a WPP through simulation. As a result, visibly heavy tails emerge in the generated power probability distribution through the use of the joint model—incorporating a detailed description of the dependence structure—in contrast with the normally distributed power yielded by the margin-based model.