Mismatching and partial shading identification in photovoltaic arrays by an artificial neural network ensemble Michel Piliougine (mpiliouginerocha@unisa.it) and Giovanni Spagnuolo (gspagnuolo@unisa.it) UNIVERSITA DEGLI STUDI DI SALERNO DIEM – Via Giovanni Paolo II 132. 84084 Fisciano (SA), Italy Date: May 22nd, 2021 Version: 1.0 Filename: dataset.mat Description: This file contains the dataset related to the article titled Mismatching and partial shading identification in photovoltaic arrays by an artificial neural network ensemble This file has been stored using the native Matlab format for its workspace (*.mat) Now, the variables stored in the file are described: Variables: (*) full_curvesOk_simulated Cell array with 5000 full IV curves simulated using the SDM, assuming uniform conditions Each IV curve is a 300 x 2 matrix (double), with column 1= voltage (V), column 2 = current (A) (all the points of the simulated IV curve, not only those around MPP) From this variable, the following splitting has been made: TRAINING => from index 1 to 3500 VALIDATION => from index 3501 to 4000 TEST => from index 4001 to 5000 (referred as TEST#1 in the paper) (*) full_curvesSh_simulated Cell array with 5000 full IV curves simulated using the SDM, assuming FAULTY conditions (steeped curve) Each IV curve is a 300 x 2 matrix (double), with column 1= voltage (V), column 2 = current (A) (all the points of the simulated IV curve, not only those around MPP) From this variable, the following splitting has been made: TRAINING => from index 1 to 3500 VALIDATION => from index 3501 to 4000 TEST => from index 4001 to 5000 (referred as TEST#1 in the paper) (*) full_curvesOk_experimental Cell array with 1000 full IV curves measured of a real PV module, assigned to be measured under normal operating conditions, and assumed to be classified as HEALTHY Each IV curve is a n x 2 matrix (double), with column 1= voltage (V), column 2 = current (A) (all the points of the simulated IV curve, not only those around MPP) where n is the number of experimental points of each IV curve, although it is around 100, but it could be different for each measured IV curve. All these curves has been used for testing TEST => from index 1 to 1000 (referred as TEST#2 in the paper) (*) full_curvesSh_experimental Cell array with 1000 full IV curves measured of a real PV module, assigned to be measured under mismatched operating conditions, and assumed to be classified as FAULTY Each IV curve is a n x 2 matrix (double), with column 1= voltage (V), column 2 = current (A) (all the points of the simulated IV curve, not only those around MPP) where n is the number of experimental points of each IV curve, although it is around 100, but it could be different for each measured IV curve. All these curves has been used for testing TEST => from index 1 to 1000 (referred as TEST#2 in the paper)