Aim: Import geodata on agricultural areas und supplemental information (climate (change), infrastructure, social, etc...) to compute potential for AV implementation.
Content:
- Create Working Geodataframe
- Get Climate data (TMYs)
- Technical and Economic modelling
- Set Parameters of AV Systems
- Shadow Simulation with optional optimization of row distance (bifacialvf)
- PV yield simulation (pvlib)
- EV simulation (pyet)
- LCOE simulation
- Multi Criteria Deciscion Making
- Create Working Geodataframe
Based on input geodata for agricultural areas and additional information, a working geodataframe (gdf) with points in a user-defined distance is created (raster-like). The information on agricultural areas and additional data is discretized and stored for every point.
- Get Climate data (TMYs)
TMYs are downloaded from PVGIS based on coordinates and saved locally as csv files. Before starting the download, existing files are checked with the working gdf: non-matching csv files are deleted and missing TMY data is downloaded. Information from the TMYs such as GHI (kWh/a), altitude (m), and information on the download ("ok", "missing climate data", "missing solar data") is saved in the working gdf.
- Technical and Economic modelling
- Set Parameters of AV Systems
Different AV systems can be defined based on the following parameters. Common parameters: "length_module", "width_module", "cap_module"
Individual parameters for each AV system: "tracking", "pvrow_azimuth", "pvrow_tilt", "pvrow_distance","pvrow_width", "pvrow_height", "bifaciality", "capex", "opex"
- Shadow Simulation with optional optimization of row distance (bifacialvf)
Based on NREL's bifacial viewfactors library (https://github.com/NREL/bifacialvf), the simulate function is modified to only perform actions that are needed to calculate irradiation on the ground (and not on the PV modules). Input variables for AV geometries have to be normalized by the PV module width (=1). The irradiation on the ground is computed for 100 points between two rows of PV modules. The irradiation values are further processed to effective PAR:
There is the possibility of applying a custom CSR to calculate the diffuse rate from the circumsolar region based on a fixed rate (instead of using the Perez model for computation). It's a bit faster, and the Perez computation, which takes DNI and DHI as inputs, results in a GHI that is not equal to the GHI in the TMY data. Optional optimization of the PV row pitch based on average DLI target value is possible (with scipy.minimize_scalar).
- PV yield simulation (pvlib)
Based on pvlib libary (https://pvlib-python.readthedocs.io/en/stable/) using the pvfactors_timeseries and pvwatts modules. Hourly AC generation is computed.
- EV simulation (pyet)
Based on pyet libary (https://github.com/pyet-org/pyet), computes the reference evapotranspiration according to FAO 56 standards.
- LCOE simulation
LCOE computation based on inhouse developed code.
- Multi Criteria Deciscion Making
The Multi-Criteria Decision-Making (MCDM) method is used in the present study to evaluate and prioritize potential areas for feasibility. Criteria like solar radiation levels, land suitability, infrastructure proximity, environmental impact, regulations, and economic factors. Each criterion is then weighted based on its importance.
With the criteria in place, potential sites are identified and scored. Techniques like Analytic Hierarchy Process (AHP) or Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) are used to calculate aggregate scores. The sites are ranked based on these scores to identify the most suitable areas for agrivoltaic development. MCDM provides a transparent and objective way to determine the feasibility of potential agrivoltaic sites, allowing for a balanced approach that considers multiple factors in the decision-making process. Respective method is a common approach applied in geospatial AV potential studies.
Potential improvements:
- More efficient optimization method for PV row pitch optimization based on effective PAR shadow value.
- Parallelization of for-loops in technical-economic computations (shadow simulation, PV yield, evapotranspiration, LCOE).
- Use NREL's bifacial viewfactors instead of pvfactors_timeseries for PV yield computation.
- Make use of light homogeneity index in optimization.
- Agregar filtros de area (clase de uso de suelo, cercania a la red electrica..)