Problem 4: Convolutional neural networks in image processing and analysis
Academic Coordinator | TBA
Business Coordinator | Manuel Quero García, CEO of Sunntics.
Specialist | Rocío Mingorance Mingorance, Head of Algorithms and Processes at Sunntics.
Scope | Convolutional neural networks (CNNs) are powerful tools for image processing and analysis, making them suitable for addressing problems such as predicting cloud coverage over a solar field from previously captured images by a cloud camera. CNNs can learn to identify spatial and temporal patterns in cloud images, enabling a certain degree of accuracy in predicting the location of a cloud over the heliostats and its density. However, the reliability of CNNs for this type of task will depend on several factors, such as the quality and quantity of available data. Evaluating the amount of data required to make accurate predictions of affected heliostat positions and the degree of impact based on cloud density can determine the success of this system’s implementation. Additionally, CNNs can be combined with temporal processing techniques, such as recurrent neural networks (RNNs) or long short-term memory (LSTM) networks, to improve time-based predictive capabilities. On the other hand, atmospheric variations, wind speed and direction, and the dynamics of cloud formation and dispersal can add uncertainty. If these variables are not adequately considered, predictions may be less reliable.