The deployment of distributed photovoltaic (PV) is accelerating worldwide. Understanding when and where PV systems will generate electricity is valuable as it affects the power balance in the grids. One way of obtaining this information is simulating the PV power production of systems detected in Remotely Sensed Data (RSD). The use of aerial imagery and machine learning models has proven effective for identifying solar energy facilities. In a Swedish research project, a Convolutional Neural Network (CNN) could identify 95% of all PV systems within a municipality. Furthermore, using Light Detection and Ranging (LiDAR) data, the orientation and area of detected PV systems can be estimated. Combining this information, with local weather and irradiance data, the historic PV power generation can be simulated. The purpose of this study is to adapt and validate a model for simulating historic decentralized PV electricity generation, based on an optimization tool developed by Becquerel Sweden, and further develop the model to simulate aggregated electricity generation on a municipality level where the individual orientation of each PV system is taken into account. The model has a temporal resolution of 1 hour and a spatial resolution of 2.5×2.5 km. A regression analysis demonstrated that the simulated generation corresponds well to the measured generation of 7 reference systems, with coefficients of determination ranging from 0.69–0.84. However, the model tends to overestimate the production compared to the measured values, with a higher total simulated production and positive mean bias errors. The correlation of the measured and generated PV power was similar, when simulating using orientations provided by the reference facility owners and LiDAR approximated orientations. Generic module parameters and an average DC/AC ratio were derived in this study, enabling simulation on a municipal level. Due to available RSD, Knivsta Municipality was the object for this study. The aggregated PV electricity generation was simulated for 2022, using both an estimation of optimal conditions and an estimation of real conditions. This was compared to the assumption that all installed AC capacity in the municipality is fed to the grid. The results show that during the highest production hour, the electricity generation resulting from estimated optimal conditions, exceeds the total installed AC capacity, while the simulation using approximated real conditions never reach the total installed AC capacity. However, the average hourly production for both scenarios, never exceeds 45% of the total installed AC capacity.
Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:uu-507263 |
Date | January 2023 |
Creators | Molin, Lisa, Ericson, Sara |
Publisher | Uppsala universitet, Byggteknik och byggd miljö |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Relation | UPTEC STS, 1650-8319 ; 23037 |
Page generated in 0.0016 seconds