Why some clean energy technologies work better together: a method to decode system synergies

Why Do Some Clean Energy Technologies Work Better Together?

Energy system models have become essential tools for planning the shift to a low-carbon economy. They can identify cost-optimal technology portfolios and trace emission reduction pathways, but they rarely explain why certain combinations of technologies outperform others. A new study published in Sustainable Energy, Grids and Networks addresses that gap with a methodology designed to surface the structural relationships between technologies, not just their individual contributions.

Two Stages-approach, One Clear Picture

The study, led by Matteo Giacomo Prina and Carlo Pelizzoni at Eurac Research's Institute for Renewable Energy, applies a two-stage approach to the Piemonte region in Italy, looking ahead to 2050 in line with European Green Deal targets. The analysis covers 17 decision variables, from solar capacity and heat pump deployment to synthetic gas production, and optimises against two objectives: minimising total annual costs and minimising CO₂ emissions.

In the first stage, the authors use the EPLANopt model to examine how six electrification technologies interact with different levels of grid decarbonisation. The analysis reveals a fundamental distinction in behaviour: heat pumps reduce CO₂ emissions regardless of how much renewable energy is already present in the grid, making them effective to deploy at any stage of the transition. Power-to-X technologies, by contrast, only produce a net environmental benefit once renewable electricity covers approximately 50% of demand. Below that threshold, deploying them can increase rather than reduce emissions. The implication for investment sequencing is direct: timing matters.

In the second stage, a deep learning surrogate model trained on thousands of EnergyPLAN simulations replaces the full energy system simulator for exploratory analysis. Operating in a fraction of the computational time of the original model, it makes it possible to systematically examine all 136 pairwise combinations of the 17 decision variables. Applying two explainable AI methods, SHAP analysis and Sobol sensitivity analysis, the authors identify solar PV capacity and synthetic gas as the dominant drivers of both cost and emissions outcomes.

The most significant pairwise synergy emerges between PV surplus generation and Power-to-Gas infrastructure: when solar production is high, having the capacity to convert that surplus into synthetic gas substantially improves overall system efficiency.

Extending the analysis to three-variable combinations reveals something invisible at the pairwise level. Battery storage acts as a critical mediator between solar generation and Power-to-Gas conversion: without sufficient battery capacity, surplus PV energy cannot be effectively channelled into the Power-to-Gas process. This is an emergent, third-order synergy, one whose effect exceeds the sum of its parts and that pairwise analysis alone would not have detected.

About This Work

The research grew from a visiting period by Matteo Prina at the University of Victoria, Canada, where he collaborated with Professor Madeleine McPherson and Mackenzie Judson, whose work on deep learning surrogate models was central to shaping the methodology. The modelling work was carried out by Carlo Pelizzoni as part of his master's thesis at Politecnico di Milano. Co-authors Giampaolo Manzolini (Politecnico di Milano), Andrea Menapace and Wolfram Sparber (Eurac Research) contributed scientific guidance throughout.

The paper is part of the ongoing work of the Overall Energy System Modelling and e-Mobility (MEM) group at the Institute for Renewable Energy, which develops energy scenario analysis tools and collaborates directly with regional governments on clean energy transition strategies. Recent applications include scenarios developed for Regione Piemonte and for South Tyrol.

The paper is open access and available at this link