A distributed photovoltaic cluster power quantification model integrating ground-based cloud image segmentation technology and high-resolution weather forecasting technology

Abstract

Abstract Under variable weather conditions, accurately predicting the power output of photovoltaic (PV) power plants using ground-based cloud image segmentation techniques is challenging due to the large amount of data and computational resources required. To address this, we propose a distributed photovoltaic cluster power prediction model that integrates ground-based cloud image segmentation with high-resolution weather forecasting technology. First, a fine-grained segmentation technique for cloud images is employed, with image preprocessing through homomorphic filtering to detect occluded cloud clusters. A threshold-based segmentation process accurately identifies cloud clusters and extracts features such as lighting intensity, transmittance, zenith distance, and cloud factor occlusion, which are used as inputs for power prediction. Additionally, a high-resolution weather forecasting system is developed to collect meteorological data relevant to distributed photovoltaic clusters and forecast their power output. To further enhance the prediction accuracy, we introduce a two-step conversion method that adjusts the predicted photovoltaic output based on temperature effects. Experimental results demonstrate that the proposed model effectively mitigates the impact of cloud cover on power output, significantly improving prediction accuracy and supporting the optimized operation of distributed photovoltaic systems.

Keywords

• Foundation cloud map
• Image segmentation
• Weather forecasting technology
• Distributed photovoltaic cluster
• Power prediction
• Temperature correction