Periodic Reporting for period 1 - TimeSmart (Timeliness of Information in Smart Grids Networks)
Okres sprawozdawczy: 2022-10-01 do 2025-03-31
At a time when vast amounts of data are being collected from the systems around us, the timeliness, or freshness, of information has become increasingly important. This is especially true in environments where data underpins critical and often time-sensitive decisions. In complex physical systems such as smart grids, the availability of real-time data is essential for maintaining operational efficiency, stability, and reliability. The TimeSmart project explored how the Age of Information (AoI), a metric that measures the time elapsed since the most recently received piece of information was generated, can enhance the monitoring and control of power system assets and improve the integration and performance of renewable energy sources.
To lay the foundation, the project developed a formal framework for evaluating timeliness, using queueing theory to model data flows and propose a taxonomy of indicators relevant to grid operations. Building on this, a simulation incorporating real-world energy data was implemented, demonstrating that prioritizing timely information can double the accuracy of five-minute-ahead energy forecasts, thereby improving short-term forecasting and, consequently, enhancing the accuracy of decisions in energy management strategies. Alongside these efforts, TimeSmart contributed architectural building blocks for future time-aware smart grids, including edge-based anomaly detection and lightweight platforms to support AI-driven decision-making at the grid edge.
To translate these insights into more practical solutions, an intelligent energy management system was developed using deep reinforcement learning and linear programming. This system, tailored for smart energy communities with energy storage in the form of community batteries, showed promising results in simulation by optimizing energy use under variable pricing conditions. Its potential was further built upon during a six-month industrial placement. Altogether, TimeSmart demonstrated how improving the freshness of information can lead to more accurate forecasting, smarter control strategies, and ultimately, more efficient and resilient energy systems.
To lay the foundation, the project developed a formal framework for evaluating timeliness, using queueing theory to model data flows and propose a taxonomy of indicators relevant to grid operations. Building on this, a simulation incorporating real-world energy data was implemented, demonstrating that prioritizing timely information can double the accuracy of five-minute-ahead energy forecasts, thereby improving short-term forecasting and, consequently, enhancing the accuracy of decisions in energy management strategies. Alongside these efforts, TimeSmart contributed architectural building blocks for future time-aware smart grids, including edge-based anomaly detection and lightweight platforms to support AI-driven decision-making at the grid edge.
To translate these insights into more practical solutions, an intelligent energy management system was developed using deep reinforcement learning and linear programming. This system, tailored for smart energy communities with energy storage in the form of community batteries, showed promising results in simulation by optimizing energy use under variable pricing conditions. Its potential was further built upon during a six-month industrial placement. Altogether, TimeSmart demonstrated how improving the freshness of information can lead to more accurate forecasting, smarter control strategies, and ultimately, more efficient and resilient energy systems.
The TimeSmart project focused on exploring the practical use of the Age of Information metric in smart grids. The goal was to assess whether this metric, which measures how recently data has been generated by sources such as smart meters, could support improved forecasting, control, and management of energy systems, especially when relying on data-driven decision-making.
Main activities and outcomes include:
1.Prioritization of Measurements collected from Smart Meters:
The project developed and tested strategies that give preference to more recent data from smart meters. The results showed that this approach can help improve short-term energy forecasting in some conditions, making forecasting models more responsive to up-to-date information.
2.Visual Representation of Data Freshness for Pattern Detection:
The project also transformed freshness-of-data traces into two-dimensional image formats using geometric encoding techniques. This made it possible to apply image-based machine learning methods to detect patterns that are not easily visible in standard time series data, suggesting a new way to incorporate information freshness into data analysis workflows.
3.Energy Control at the Household Level:
A combined control method was developed that used reinforcement learning from multiple agents together with mathematical optimization. This method was tested in a simulated home energy environment, where it managed energy storage systems under variable pricing conditions. It helped reduce energy costs while maintaining consistent access to electricity for the user.
4.Coordination of Energy Use Across Multiple Households:
The cooperative reinforcement learning-based control approach was extended to small energy community settings, where several homes with their own photovoltaic system and batteries shared locally generated and stored energy. Simulations showed that the system could lower peak electricity demand from external sources and balance energy use across the community more effectively. The freshness of the input data played an important role in how well these control agents coordinated.
In summary, the TimeSmart project explored several ways to apply the Age of Information metric in smart grids, focusing on forecasting and management of smart energy communities equipped with renewable energy sources and energy storage systems. The methods developed were tested in simulation and provide early evidence that data freshness can support better short-term decision-making in time-sensitive smart grids.
Main activities and outcomes include:
1.Prioritization of Measurements collected from Smart Meters:
The project developed and tested strategies that give preference to more recent data from smart meters. The results showed that this approach can help improve short-term energy forecasting in some conditions, making forecasting models more responsive to up-to-date information.
2.Visual Representation of Data Freshness for Pattern Detection:
The project also transformed freshness-of-data traces into two-dimensional image formats using geometric encoding techniques. This made it possible to apply image-based machine learning methods to detect patterns that are not easily visible in standard time series data, suggesting a new way to incorporate information freshness into data analysis workflows.
3.Energy Control at the Household Level:
A combined control method was developed that used reinforcement learning from multiple agents together with mathematical optimization. This method was tested in a simulated home energy environment, where it managed energy storage systems under variable pricing conditions. It helped reduce energy costs while maintaining consistent access to electricity for the user.
4.Coordination of Energy Use Across Multiple Households:
The cooperative reinforcement learning-based control approach was extended to small energy community settings, where several homes with their own photovoltaic system and batteries shared locally generated and stored energy. Simulations showed that the system could lower peak electricity demand from external sources and balance energy use across the community more effectively. The freshness of the input data played an important role in how well these control agents coordinated.
In summary, the TimeSmart project explored several ways to apply the Age of Information metric in smart grids, focusing on forecasting and management of smart energy communities equipped with renewable energy sources and energy storage systems. The methods developed were tested in simulation and provide early evidence that data freshness can support better short-term decision-making in time-sensitive smart grids.
The TimeSmart project advanced the state of the art by demonstrating the practical use of the Age of Information metric in smart grids, showing how the freshness of measurement data can improve short-term energy forecasting. The project also contributed by developing control strategies based on cooperative learning methods, which support more cost-effective management of energy storage systems at both the household and energy community levels, helping to reduce electricity costs for users. Together, these findings suggest that the Age of Information metric can be a valuable tool for enhancing the responsiveness and efficiency of distributed energy systems.
To support further uptake, all code and datasets developed during the project have been released as open-source under FAIR principles. Future work may focus on expanding these approaches to larger testing environments, integrating them with edge-to-cloud data processing systems, and aligning with European Union policy frameworks, which aim to improve the performance of renewable energy systems within local energy communities.
To support further uptake, all code and datasets developed during the project have been released as open-source under FAIR principles. Future work may focus on expanding these approaches to larger testing environments, integrating them with edge-to-cloud data processing systems, and aligning with European Union policy frameworks, which aim to improve the performance of renewable energy systems within local energy communities.