Samples of wind power and electricity prices used in the manuscript "Risk-averse stochastic programming vs. adaptive robust optimization: a virtual power plant application"

Description of the files with the wind power and electricity prices scenarios used in the manuscript: "Risk-averse stochastic programming vs. adaptive robust optimization: a virtual power plant application" by Ricardo M. Lima, Antonio J. Conejo, Loic Giraldi, Olivier Le Maitre, Ibrahim Hoteit, and Omar Knio, submitted to the INFORMS Journal on Computing. This file describes the wind power and electricity prices for two weeks used in the manuscript above. Please refer to the manuscripts above for a detailed description of the data sources, sampling process, and its characteristics. The sample average approximation methodology involves two stages: 1) optimization stage; 2) bound estimation stage. Data for the the stochastic programming approach The optimization stage uses sample sizes of N=10, 50, 100, 500, and 5000; and for each sample size, M = 30 optimization replications are performed. This data is available in the following files: wp_week{X}_N{Y}_M{Z}.csv pp_week{X}_N{Y}_M{Z}.csv where X:={1,2} is the week number Y:={10,50,100,500,5000} is the sample size Z:={1,...,30} is the replication number The bound estimation stage, the lower bound on the true optimal objective function value is estimated in two steps: 1) for each distinct first-stage solution obtained from the optimization replications, a lower bound is estimated using T=30 samples of size N=25,000, and 2) the first-stage solution with the best lower bound from the previous step is selected and a new lower bound is estimated using S=30 samples of size N=25,000. The data is available in the following files wp_week{X}_N{Y}_T{Z}.csv pp_week{X}_N{Y}_T{Z}.csv wp_week{X}_N{Y}_S{Z}.csv pp_week{X}_N{Y}_S{Z}.csv where X:={1,2} is the week number Y:={25000} is the sample size Z:={1,...,30} is the replication number Data for the adaptive robust optimization approach The optimization stage uses uncertainty sets defined in these files: pp_week{X}_pf.csv pp_week{X}_95ci.csv wp_week{X}_ensemble.csv where X:={1,2} is the week number The lower bound estimation stage requires only the second step above, using S = 30 samples of size N=25,000. wp_week{X}_N{Y}_S{Z}.csv pp_week{X}_N{Y}_S{Z}.csv where X:={1,2} is the week number Y:={25000} is the sample size Z:={1,...,30} is the replication number