Simplified, Robust Method for Cloud Predictions for Photovoltaic Estimations

Photovoltaics is becoming a critical component in the provision of electric energy to most power grids. However, its output is highly dependent on the position and characteristics of clouds moving over the photovoltaic farm which reduce incident sun irradiation. Knowledge of the dynamics of these clouds, such as size and velocity can be used to optimize the use of such power sources in their connected grids. Cloud behaviours can be relatively fast compared to the system dynamics of a grid, and the variability of such power into the grid and its problematic effect is well known, including suboptimal and unstable grid performance. This research, which is ongoing, focuses on cloud-dynamics prediction in real-time for optimal grid control. The work uses simplified, first principle image and signal processing deliberately avoiding neural networks and related issues such as training and data validation. The system developed predicts cloud positions of well defined cloud shapes within acceptable errors and given probabilities which can be used for predicting photovoltaic power for grid control and optimization.