Modeling multivariate intraday forecast update processes for wind power
针对日内市场连续交易中风电预测不断更新的特点，提出了一个基于copula的多变量随机过程模型，用于刻画预测更新的相关性，帮助制定电力交易策略。
With the on-going expansion of renewable energy generation, short-term trading, notably in intraday markets, becomes increasingly relevant to cope with forecast updates for renewable infeeds. In this context, we develop a multivariate model of wind forecasting trajectories in order to support power trading strategies under continuous trading with repeated updates of wind forecasts. Thereby, we consider the correlations of forecast changes for subsequent delivery periods as these are e.g. relevant for storage operation and marketing. Based on theoretical hypotheses on properties of forecast trajectories, we propose a multivariate stochastic process based on copulas applied to forecast updates. The model is applied to wind forecast data of the French TSO and compared to alternative model specifications proposed in the literature. The newly developed approach provides a balanced performance across three distinct metrics — measuring overall forecast performance as well as the ability to replicate dependencies along and between forecast trajectories.