Optimization and data aggregation for net-zero power systems

Project Information

NetZero-Opt

Grant agreement ID: 101116212

DOI

10.3030/101116212

EC signature date 28 August 2023

Start date 1 January 2024

End date 31 December 2028

Funded under

• European Research Council (ERC)

Total cost

€ 1 499 888,00

EU contribution

€ 1 499 888,00

Coordinated by

TECHNISCHE UNIVERSITAET GRAZ
Austria

Periodic Reporting for period 1 - NetZero-Opt (Optimization and data aggregation for net-zero power systems)

Reporting period: 2024-01-01 to 2026-06-30

Summary of the context and overall objectives of the project

The European Union has set binding targets to cut greenhouse gas emissions by 55 % by 2030 and reach climate neutrality (net-zero emissions) by 2050 under the Green Deal and Climate Law. Achieving this requires transforming Europe’s energy systems, especially the power system, which must ensure reliability, affordability, and sustainability across time horizons from minutes to decades. Optimization models are central to this, guiding operation, investment, and long-term transition decisions.

A major obstacle is computational complexity. Modern models use large time-series datasets—demand, renewable generation, weather—and must capture both short-term constraints and long-term dynamics. Detailed models often become too large or slow to solve, limiting practical use.
Time series aggregation (TSA) reduces data to make optimization feasible. Existing TSA, however, often ignores interacting short- and long-term dynamics and relies on heuristics assuming input accuracy ensures correct outputs. Resulting decision errors remain poorly understood.

NetZero-Opt addresses these limitations by uncovering the relationship between input aggregation and output errors, developing a theoretically grounded, structure-aware TSA framework that accounts for model structure and dynamics. By focusing on decision relevance rather than data similarity, it enables models that are efficient and reliable.

The impact is twofold. Scientifically, it provides a principled, theory-driven foundation for data aggregation. Societally, it supports scalable, trustworthy analyses of net-zero pathways, informing better planning and policy, and removing a key computational bottleneck.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The NetZero-Opt project builds on a systematic exploration of time series aggregation (TSA) for optimization models, addressing large-scale, complex power systems. Check out the project website: https://www.tugraz.at/en/institutes/iee/research/current-projects/netzero-opt

Proof of concept: a posteriori structure-aware TSA for linear programs
The project originates from Wogrin (2023), which shows for linear programs that explicitly accounting for model structure (Simplex basis) enables substantial data aggregation without loss of solution quality. This work established the theoretical possibility of exact aggregation in the objective function and outputs, providing the conceptual foundation for NetZero-Opt.

Extension to network and ramping constraints
Building on this foundation, Cardona-Vasquez et al. (2024) extended TSA to models with network and ramping constraints, common in electricity and gas systems. The study demonstrated that even with time-linking constraints—such as thermal unit ramping—perfect aggregation is possible. These results, requiring full model solutions, served as proofs of concept and established the framework for structure-aware TSA.

Machine learning for practical TSA without full model solutions
A subsequent step addressed the need for full outputs. In Cardona-Vasquez et al. (2025), machine learning was used to predict the activation of time-linking constraints. Identifying inactive constraints allowed partitioning the system matrix into independent subproblems without output errors and enabled parallelization, achieving 300× speed-ups on the Austrian grid.

Introducing performance bounds for aggregated models
Another key result is the development of bounded TSA methods for optimization models, including MILP formulations such as Virtual Power Plant (VPP) investment planning. Santosuosso et al. (2025) formally proved how to construct aggregated models that provide both lower and upper bounds on the full model solution, ensuring performance guarantees and feasibility throughout iterative procedures—beyond the heuristic nature of existing TSA approaches.

Towards exact aggregation with energy storage constraints
In subsequent work, Klatzer et al. (2025) and Santosuosso et al. (2025b) achieved exact TSA for models with energy storage time-coupling constraints. By exploiting active constraint sets, dual information, and machine learning for period classification and clustering, the methodology enables accurate aggregation in co-scheduling problems with variable renewable energy and storage.

Extension to spatial aggregation and model predictive control
To address challenges from high spatial resolution, NetZero-Opt extended its framework to spatial aggregation. Stöckl et al. (2025) developed aggregation methods based on a Network Congestion Price metric that preserve the impact of nodal injections on line congestion, outperforming standard Locational Marginal Price–based approaches. The project also released the open-source Python Network Partitioning & Aggregation Package (NPAP) to support practical application. Moreover, in an initial publication (Santosuosso et al 2025c) we show that structure-aware aggregation can be extended to other fields such as model predictive control.

Summary of achievements
NetZero-Opt established a comprehensive framework for time series and spatial aggregation of large-scale energy system optimization models, including:

1. Structure-aware TSA proofs of concept
2. Extensions to network and ramping constraints
3. Practical TSA via machine learning without full model solutions
4. Performance-bounded aggregation for LP and MILP models
5. Exact TSA for storage-coupled systems
6. Novel spatial aggregation methods and an open-source software package (NPAP)
7. Extension to model predictive control

These results advance the state of the art by combining theoretical rigor, computational efficiency, and practical applicability for scalable energy system optimization.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

NetZero-Opt advances the state of the art by establishing a performance-guaranteed paradigm for aggregation in optimization, moving beyond heuristic and data-driven approaches. The project demonstrates that aggregation can be exact or bounded while explicitly accounting for network constraints, ramping limits, storage dynamics, and mixed-integer formulations—capabilities largely absent from existing TSA methods.

A central contribution is the shift from input-based aggregation to decision-relevant accuracy, supported by theoretical guarantees on outputs. Machine learning enables aggregation without full model solutions, improving scalability and enabling parallelization.

Beyond the temporal dimension, NetZero-Opt extends aggregation to spatial complexity through novel congestion-sensitive methods and provides an open-source software tool to support adoption. These results remove bottlenecks limiting detailed optimization in energy system planning. Further uptake will benefit from research on long-term storage, integration into iterative algorithms, and broader adoption of the project’s open-source tools.

Overview of results
Up until now, the project has delivered:
1. A structure-aware, theoretically grounded TSA framework
2. Exact aggregation for linear models with network, ramping, and storage constraints
3. Performance-guaranteed aggregation with bounds for mixed-integer models
4. Machine learning-informed TSA enabling substantial computational speed-ups
5. Novel spatial aggregation methods and an open-source Python package (NPAP)
6. Extension to model predictive control