Cloudiness Forecasting Using Satellite Data Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting cloudiness, especially from satellite image analysis, are prone to precision loss and sensitivity to aberrant data due to limited data history and spatial overlap, leading to unreliable short-term forecasting of solar radiation, which affects electricity production optimization in solar energy systems.
Innovation Solution
A method combining satellite image processing with statistical modeling using a generalized linear model and spatial analysis to select and process data from a larger history of cloudiness instants, incorporating wind characteristics and ground measurements to improve robustness and accuracy of cloudiness forecasting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If satellite image analysis is used for cloud cover forecasting, then spatial coverage and data quantity are improved, but measurement precision and reliability deteriorate due to sensitivity to outliers and data defects
Solution Approach 1:
The patent extracts and removes defective or aberrant satellite images from the analysis dataset through quality control procedures. By identifying and excluding images with defects (such as cloud contamination, atmospheric interference, or sensor errors), the method prevents these outliers from degrading the overall estimation precision while maintaining the benefits of using multiple satellite images for cloud cover forecasting.
Solution Approach 2:
The patent performs preliminary quality assessment and filtering of satellite images before they are used in cloud cover estimation. By pre-screening images for defects and aberrations, and by establishing quality thresholds in advance, the method ensures that only reliable data contribute to the final estimation, thereby maintaining precision while utilizing large quantities of satellite data.
2Device complexity
If a limited data history is used in satellite image analysis, then processing complexity is reduced, but reliability deteriorates due to increased sensitivity to outliers and data defects
Solution Approach 1:
The patent extracts and removes defective or aberrant satellite images from the analysis dataset through quality control procedures. By identifying and excluding images with defects (such as cloud contamination, atmospheric interference, or sensor errors), the method prevents these outliers from degrading the overall estimation precision while maintaining the benefits of using multiple satellite images for cloud cover forecasting.
Solution Approach 2:
The patent implements feedback mechanisms through quality control metrics that continuously assess the reliability of satellite data. By monitoring data quality indicators and adjusting the inclusion/exclusion of data points based on observed performance, the method dynamically balances the use of historical data quantity against reliability concerns, ensuring robust forecasts even with limited data histories.
3Reliability
If statistical modeling with large historical data is used, then robustness to outliers is improved, but input data selection complexity increases when data must be selected for each application
Solution Approach 1:
The patent performs preliminary quality assessment and filtering of satellite images before they are used in cloud cover estimation. By pre-screening images for defects and aberrations, and by establishing quality thresholds in advance, the method ensures that only reliable data contribute to the final estimation, thereby maintaining precision while utilizing large quantities of satellite data.
Solution Approach 2:
The patent adjusts data selection parameters and quality thresholds based on the specific forecasting application and environmental conditions. By dynamically modifying parameters such as the temporal window size, spatial resolution, and quality acceptance criteria, the method optimizes the balance between utilizing sufficient historical data for robustness and managing the complexity of data selection for each specific forecast application.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method and a device providing cloudiness forecasts by statistical processing of satellite data which can be combined with data measured on the ground. The method includes collecting (11) satellite data on cloudiness at n times {T1,...,Tn} (n>1), analysing (12 and 13) the movement (X) of the cloudiness values between two times, recovering (15) a history of data (DATA_SELEC) selected relative to (X), and then estimating (16) the cloudiness at a time (TH) later than the other times by applying a generalised linear model (MOD_G) to the data (DATA_SELEC). Data from sensors placed on the ground, communicating and self-contained, can advantageously be added to the input of (MOD_G). The device comprising said method in the form of a computer program product is specifically intended for forecasting the electricity generated by solar conversion systems.