Validation of a PV generation model for simulation of wide area aggregated distributed PV power generation that takes into individual systems locaiton and orientation into account

• Understanding the photovoltaic (PV) power generation's temporal and spatial patterns is vital for grid balancing. This study aims to validate a simulation model for historical decentralized PV power generation, extending it to encompass the unique orientation of all PV systems within the Swedish municipality Knivsta. In a previous research project, a Convolutional Neural Network exhibited a 95% accuracy of identifying PV systems within Knivsta. In this project, using Light Detection and Ranging data, the orientation and area of detected PV systems was estimated. By combining this information with local weather and irradiance data, historical PV power generation was simulated. The regression analysis demonstrates strong correspondence between simulated and measured hourly generation for six reference systems, with coefficients of determination between 0.69–0.83. This study derives generic module parameters based on installation year and an average DC-to-AC ratio, enabling municipal-level simulations. Simulations for 2022, considering one scenario with optimal orientation for all PV systems and one scenario with derived real-condition orientations, reveal a smoothing effect in the daily pattern of aggregated PV generation, if considering real orientations. At the peak hour, power generation was found to be 10% lower when considering individual orientations compared to assuming optimal orientation across all facilities.

Ämnesord

TEKNIK OCH TEKNOLOGIER  -- Samhällsbyggnadsteknik (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Civil Engineering (hsv//eng)

Publikations- och innehållstyp

vet (ämneskategori)

kon (ämneskategori)