Feature-Driven Improvement of Renewable Energy Forecasting and Trading

Abstract

Inspired from recent insights into the common ground of machine learning, optimization and decision-making, this paper proposes an easy-to-implement, but effective procedure to enhance both the quality of renewable energy forecasts and the competitive edge of renewable energy producers in electricity markets with a dual-price settlement of imbalances. The quality and economic gains brought by the proposed procedure essentially stem from the utilization of valuable predictors (also known as features) in a data-driven newsvendor model that renders a computationally inexpensive linear program. We illustrate the proposed procedure and numerically assess its benefits on a realistic case study that considers the aggregate wind power production in the Danish DK1 bidding zone as the variable to be predicted and traded. Within this context, our procedure leverages, among others, spatial information in the form of wind power forecasts issued by transmission system operators (TSO) in surrounding bidding zones and publicly available in online platforms. We show that our method is able to improve the quality of the wind power forecast issued by the Danish TSO by several percentage points (when measured in terms of the mean absolute or the root mean square error) and to significantly reduce the balancing costs incurred by the wind power producer.Index Terms-Electricity markets, Renewable energy forecasting and trading, Wind power, Optimization, Machine Learning. NOMENCLATURE A. Sets and Indicest Index of time periods. j Index of features. T Training set. T Test set. B. Parameters E Maximum hourly wind energy production (MWh). λ B t Balancing market price at hour t (e/MWh). λ D t Day-ahead market price at hour t (e/MWh). λ − t Upward regulation price in the balancing market at hour t (e/MWh). λ + t Downward regulation price in the balancing market at hour t (e/MWh). ψ − t Marginal opportunity cost for underproduction at hour t (e/MWh). ψ + t Marginal opportunity cost for overproduction at hour t (e/MWh). E t Actual wind energy produced at hour t (MWh). M. A. Muñoz, J. M. Morales (corresponding author) and S. Pineda are with the research group OASYS at

Prediction Markets for Probabilistic Forecasting of Renewable Energy Sources

“…where + and − are imbalance unit costs for the positive and negative imbalances, respectively. These prices can be considered as a certain proportion of : + = (1 − + ) and − = (1 + − ) , incurring penalties in underproduction situations and payments with a rate lower than the day-ahead market clearing price for the overproduction [13].…”

“…The optimal day-ahead bid value for a stochastic res is the of ( 11) in order to minimise the imbalance costs. As shown in [13] and [44], a specific percentile of the predictive cdf of the res output, which is a function of imbalance prices in the electricity market and corresponds to the optimal dayahead bid value, as expressed in:…”

“…Extensive research exists on this topic in the literature, for a detailed classification and review, we refer to [8], [9] and [10]. Here, within the scope of this paper, it suffices to note that these methods can be categorised into two groups: parametric methods assume a pre-defined shape for the distribution of forecast errors, such as Gaussian or beta ( [11] and [12]) and estimate the relevant parameters, while nonparametric methods obtain empirical probability distributions from the historical data ( [13] and [14]) without any prior assumption on the shape of errors distribution.…”

Short‐term load forecasting based on strategies of daily load classification and feature set reconstruction

“…The number of filters in each residual block is [64,32,24,16,12,8,8]. And the dilation factor is [1,2,4,8,16,32,64]. The kernel size is 2.…”

A Value-Oriented Price Forecasting Approach to Optimize Trading of Renewable Generation

“…The same work also introduces the coefficient of prescriptiveness, a unitless metric analogous to the R 2 coefficient in regression, which is used to assess the relative performance of prescribed decisions. From an applications standpoint, [19] proposes a data-driven approach to increase forecast accuracy and trading value in the presence of deterministic forecasts of wind power.…”