Estimating the Output of Behind the Meter Solar Farms by Breaking Irradiance Data into its Diffuse and Direct Components

As more behind the meter solar farms are installed onto the power grid, the true load on the power grid becomes more hidden to the utility because the meters only read the net difference between the native load and the solar generation. This becomes problematic for grid planning since the grid needs to be ready to handle the full native load in the case of a hot and cloudy summer day when load is very high and solar generation is low. This study breaks down solar irradiance into the diffuse components from the ground and sky, and the direct (beam) irradiance. These components are then combined using the Liu-Jordan model to estimate the solar irradiance on a tilted surface. This method was then applied to data from a weather station in an area in ComEd’s service territory to estimate the solar panel output of a local 2MW metered solar farm. When comparing the predicted generation and measured generation from this solar farm, it was determined that their existed inconsistencies within the data set. After reducing the size of the data set to remove potentially poor data, this method estimated solar production with an R
²
value of 0.900 with an average absolute value error of 148kW. Based on these findings, this methodology had produced efficient results and can also be used to determine when a solar farm is not producing as expected.