Clustering techniques for photovoltaic system simulations under nonuniform irradiance conditions

Accurate and fast simulations of photovoltaic systems are essential for streamlining system design and for correctly assessing their economic merit and impact on the electric grid. Typically, simulations of these systems are done over a period of one year with hourly resolution, and require taking into account the different operating conditions of each solar panel, e.g., incident irradiance, as well as circuit constraints, e.g., Kirchhoff´s laws. Under nonuniform irradiance conditions, a typical consequence of shading, differences in panel operating conditions significantly increase simulation complexity and time. This is because such simulations require solving circuits of panels having potentially many different current-voltage characteristics. In this work, techniques are explored for reducing model complexity in such a way that systems operating under nonuniform irradiance conditions can be simulated fast and accurately. The techniques explored are based on identifying a suitable number of panel groups based on panel energy dispersion, clustering panels based on sub-sampled irradiance conditions into the identified number of groups, and reducing coupling among the resulting system components. The techniques have been tested using Aurora Solar´s module-level simulation engine on four systems that operate under nonuniform irradiance conditions. The results obtained suggest that significant simulation speedups can be obtained while maintaining adequate levels of accuracy.