Menu Close

Which gridded irradiance data is best for modelling photovoltaic power production in Germany?

Mod­el esti­mates of expect­ed pho­to­volta­ic (PV) pow­er pro­duc­tion rely on accu­rate irra­di­ance data. Reanaly­sis and satel­lite prod­ucts freely pro­vide irra­di­ance data with a high tem­po­ral and spa­tial res­o­lu­tion includ­ing loca­tions for which no ground-based mea­sure­ments are avail­able. We assess dif­fer­ences in such grid­ded irra­di­ance data and quan­ti­fy the sub­se­quent bias prop­a­ga­tion from indi­vid­ual radi­a­tion com­po­nents to capac­i­ty fac­tors in a con­tem­po­rary PV mod­el. PV pow­er pro­duc­tion is sim­u­lat­ed based on four reanaly­sis (ERA5, COSMO-REA6, COS­MO-REA6pp, COSMO-REA2) and three satel­lite prod­ucts (CAMS, SARAH‑2, CERES Syn1Deg). The results are com­pared against sim­u­la­tions using mea­sure­ments from 30 weath­er sta­tions of the Ger­man Weath­er Ser­vice. We com­pute met­rics char­ac­ter­iz­ing bias­es in sea­son­al and annu­al means, day-to-day vari­abil­i­ty and extremes in PV pow­er. Our results high­light a bias of −1.4% (COSMO-REA6) to +8.2% (ERA5) in annu­al and spa­tial means of PV pow­er pro­duc­tion for Ger­many. No sin­gle data set is best in all met­rics, although SARAH‑2 and the post­processed COSMO-REA6 data (COS­MO-REA6pp) out­per­form the oth­er prod­ucts for many met­rics. SARAH‑2 yields good results in sum­mer, but over­es­ti­mates PV out­put in win­ter by 16% aver­aged across all sta­tions. COS­MO-REA6pp rep­re­sents day-to-day vari­abil­i­ty in the PV pow­er pro­duc­tion of a sim­u­lat­ed PV fleet best and has a par­tic­u­lar­ly small bias of 0.5% in annu­al means. This is at least in parts due to com­pen­sat­ing bias­es in local and sea­son­al means. Our results imply that grid­ded irra­di­ance data should be used with cau­tion for site assess­ments and ide­al­ly be com­ple­ment­ed by local mea­sure­ments.

Dar­raghKen­ny, StephanieFiedler (2022) in Solar Ener­gy, Vol­ume 232, 15 Jan­u­ary 2022, Pages 444–458

Link to arti­cle

Publications, Research