An ultra-short-term power prediction model based on machine vision for distributed photovoltaic system

Distributed photovoltaic(PV) system is easily affected by the cloud cluster moving in the sky because of its small scale. The instantaneous shelter caused by the moving cloud cluster may lead to the output power of photovoltaic system fluctuation violently. The cloud cluster monitoring device was designed, which aims to track the solar trajectory and take photos of the cloud cluster. The centroid position feature model and shape feature model were established based on image-based processing algorithms. They can forecast the position and shape of cloud cluster in the near future. And an ultra-short-term power prediction model based on machine vision for distributed photovoltaic system was established. Simulation results show that the established model can track the position of cloud cluster in the sky, and predict the shape-to-be of cloud cluster.