The ever-growing penetration of cost-effective photovoltaic (PV) panels within the distribution grid requires a robust and efficient method for PV system monitoring. Especially, the geographical proximity of PV panels can play an important role in lowering the dimension of measurements required for full system observability. Furthermore, the direct impact of variable cloud formation and uncertain propagation necessitates the development and validation of a spatiotemporal model. Accordingly, this study presents the modelling and validation of the spatiotemporal variability of solar power indices at 1 minute resolution for the scale of a residential neighbourhood. The spatiotemporal model is then applied to a Multi-Rate and Event-DRIven Kalman Kriging (MREDRIKK) filter to dynamically estimate behind-the-meter PV generation. The Kriging step exploits spatial correlations to estimate PV power output at locations from where measurements are unobserved. The multi-rate feature of the MREDRIKK filter enables the sampling of measurements at a rate much lower than the temporal dynamics of the associated states. A comprehensive study is undertaken to investigate the effect of multi-rate and event-driven measurement updates on the performance of the MREDRIKK filter. In addition, the superior performance of MREDRIKK filter is represented as compared to the persistence method irrespective of the observation size.

No abstract

No abstract