Solar Forecasting for Isolated Microgrids

Powerful computers, large databases, and historical weather-forecast reach-backs are often associated with successful, contemporary, solar power forecasting. What if you had only local, onsite input? Could you optimize a hybrid microgrid?

Optimizing the integration of traditional (generator/battery) and non-traditional (solar) resources in an isolated hybridized power grid is the framework for this research. This presentation begins by defining a hybrid microgrid design, along with investigative metrics. We then examine several atmospheric models used to provide critical atmospheric intelligence used by the optimization strategies. The strengths and shortfalls of each model are quantified through grid metrics and distinctive model characteristics. The impact of meteorological forecast frequency on fuel consumption and energy storage for the isolated electrical grid, is examined. We also explore management algorithms which allowed fuel consumption to be invariant with respect to forecast frequency and atmospheric conditions. These showed increased storage requirements. We found that exploiting atmospheric input allowed for the routing of PV directly to loads, which avoided losses associated with the movement of stored energy. When future loads exceeded generator capacity, storage had to be used. Maintaining peak generator efficiency was a function of storage, but better managed with local atmospheric assessment. We concluded that an isolated grid operated using weather and load forecast-based energy-management strategies, forecast periods, and qualities should be considered when sizing grid storage capacity. Improper consideration can result in sub-optimal energy management.