Search papers, labs, and topics across Lattice.
This paper introduces a two-stage photovoltaic (PV) forecasting model that separates weather prediction from plant-specific characteristics to improve forecast accuracy and error distribution analysis. The approach uses satellite-based weather observations as an intermediate layer, feeding them into an ensemble of neural networks trained on historical power output data to model plant characteristics. Results demonstrate that using weather forecasts instead of satellite observations increases the mean absolute error by 11% and 68% for two PV systems, highlighting the impact of weather forecast inaccuracies on PV prediction.
Weather forecast inaccuracies can inflate photovoltaic (PV) forecasting errors by as much as 68%, underscoring the importance of disentangling weather prediction from plant-specific factors.
Several energy management applications rely on accurate photovoltaic generation forecasts. Common metrics like mean absolute error or root-mean-square error, omit error-distribution details needed for stochastic optimization. In addition, several approaches use weather forecasts as inputs without analyzing the source of the prediction error. To overcome this gap, we decompose forecasting into a weather forecast model for environmental parameters such as solar irradiance and temperature and a plant characteristic model that captures site-specific parameters like panel orientation, temperature influence, or regular shading. Satellite-based weather observation serves as an intermediate layer. We analyze the error distribution of the high-resolution rapid-refresh numerical weather prediction model that covers the United States as a black-box model for weather forecasting and train an ensemble of neural networks on historical power output data for the plant characteristic model. Results show mean absolute error increases by 11% and 68% for two selected photovoltaic systems when using weather forecasts instead of satellite-based ground-truth weather observations as a perfect forecast. The generalized hyperbolic and Student's t distributions adequately fit the forecast errors across lead times.
