Search

While the large-scale deployment of photovoltaics (PV) for generating electricity plays an important role to mitigate global warming, the variability of PV output power poses challenges in grid management. Typically, the PV output power is dependent on various meteorological variables at the PV site. In this paper, we present a systematic approach to perform an analysis on different meteorological variables, namely temperature, dew point temperature, relative humidity, visibility, air pressure, wind speed, cloud cover, wind bearing and precipitation, and assess their impact on PV output power estimation. The study uses three years of input meteorological data and PV output power data from multiple prosumers in two case studies, one in the U.S. and one in the Netherlands. The analysis covers the correlation and interdependence among the meteorological variables. Then, by using machine learning-based regression methods, we identify the primary meteorological variables for PV output power estimation. Finally, the paper concludes that the impact of using a lower-dimensional subspace of meteorological variables per location, as input for the regression methods, results in a similar estimation accuracy in the two case studies. • An analysis of the correlation and interdependence of 9 different meteorological variables in two case studies. • A dimensionality reduction of the input meteorological variables is presented. • Several machine learning regression methods are evaluated for estimating PV output power. • The importance and ranking of the variables depend on the climate of the area of study. • A lower-dimension subspace of meteorological variables can result in a slightly similar estimation accuracy. [ABSTRACT FROM AUTHOR]

This paper studies the use of conformal prediction (CP), an emerging probabilistic forecasting method, for day-ahead photovoltaic power predictions to enhance participation in electricity markets. First, machine learning models are used to construct point predictions. Thereafter, several variants of CP are implemented to quantify the uncertainty of those predictions by creating CP intervals and cumulative distribution functions. Optimal quantity bids for the electricity market are estimated using several bidding strategies under uncertainty, namely: trust-the-forecast, worst-case, Newsvendor and expected utility maximization (EUM). Results show that CP in combination with k-nearest neighbors and/or Mondrian binning outperforms its corresponding linear quantile regressors. Using CP in combination with certain bidding strategies can yield high profit with minimal energy imbalance. In concrete, using conformal predictive systems with k-nearest neighbors and Mondrian binning after random forest regression yields the best profit and imbalance regardless of the decision-making strategy. Combining this uncertainty quantification method with the EUM strategy with conditional value at risk (CVaR) can yield up to 93% of the potential profit with minimal energy imbalance. • A novel framework using conformal prediction to aid PV power suppliers in decision-making for the day-ahead electricity market. • Various variants of CP with different market bidding strategies were developed and evaluated. • CP methods outperformed linear quantile regression models. • CPS with KNN and Mondrian binning showed superior performance in profit and imbalance across considered bidding strategies. • Risk aversion considerations may differently influence the selection of bidding strategies. [ABSTRACT FROM AUTHOR]