EOSC-SIESTA has been designed to address one of the main challenges in open science: enabling the analysis and reuse of sensitive data without compromising privacy. To do so, the project develops tools and services that allow researchers and organisations to work with real datasets within secure, cloud-based environments, combining accessibility with robust privacy preserving mechanisms.
Within this framework, UC3, focused on the energy domain represents a practical example of how these principles can be applied. This use case aims to optimise the management of solar energy systems through data driven decision making, while ensuring that sensitive information remains protected throughout the entire process.
The scenario is based on solar installations located in southern Spain, where the central challenge lies in deciding how to manage the energy generated on a daily basis. This involves determining when it is more efficient to store energy in batteries, when to sell it to the grid, and when to purchase electricity, all with the objective of reducing overall costs.
To address this, UC3 relies on predictive models that estimate both energy production and consumption. On the one hand, the system forecasts the amount of energy that will be generated over the next 24 hours by combining weather data, such as solar radiation, with historical production records. On the other hand, it predicts household consumption patterns based on past usage. These two elements are then integrated with real-time electricity market prices, allowing the system to provide recommendations on how to manage energy in the most efficient way.
A key aspect of this use case is the nature of the data involved. While production data can generally be handled without major concerns, consumption data reflects the behaviour of individuals within a household, making it highly sensitive. This introduces an additional layer of complexity, as it requires ensuring that such data is handled securely at all times.
EOSC-SIESTA addresses this challenge through the use of trusted execution environments, which enable models to be trained and executed without exposing the underlying data. In this way, sensitive information remains within controlled environments, reducing the risk of data leakage while still allowing meaningful analysis to take place.
In parallel, UC3 also explores anonymisation strategies that complement these secure infrastructures. One of the approaches being developed involves the use of computer vision techniques to identify and cluster nearby solar panel installations. By grouping installations within a defined geographical range, the system reduces the risk of identifying individual households, which is particularly relevant when preparing datasets for publication or for training anonymised machine learning models.
From the user perspective, the UC3 service can be accessed through the EOSC-SIESTA dashboard, where the full workflow is integrated into a single environment. Users can locate the service within the catalog, launch it directly, configure their access to the working environment and interact with it through an interface that allows them to run and adapt the available tools according to their needs.
To facilitate this process, EOSC-SIESTA provides a step by step tutorial that guides users through the deployment and execution of the UC3 service. This tutorial explains how to access the platform, launch the tool and interact with the environment, offering a practical entry point for those who want to understand how the use case operates in practice.
In this video, we navigate to the service catalog and instantiate the uc3-energy-forecasting service. Upon launching the service, the system automatically creates a container for us. Since this service is a FastAPI-based API, we can use our own historical consumption or generation time-series data to generate forecasts by simply calling the api.
In addition, the UC3 use case is supported by a set of complementary resources that allow users to explore the technical aspects in more detail. The implementation is available through its corresponding GitHub repository, where the code and development components can be accessed, while the official documentation provides further guidance on configuration, usage and integration within the EOSC-SIESTA ecosystem. These materials, together with demonstration content, offer a complete overview of how the service is deployed and how it can be reused or adapted in different contexts.