Chiara Fusar Bassini, a PhD candidate at the Hertie School and former renewable energy consultant at enervis energy advisors GmbH, works at the intersection of machine learning, energy-market analysis, and system design. Her work focuses on how real-world data can improve the understanding of electricity markets and the constraints shaping the energy transition.

In a conversation hosted by Net Zero Compare, she shared insights into how energy-market models perform in practice, where data-driven approaches add value, and what limitations practitioners should be aware of when working with analytical tools.

🎥 Watch the Full Conversation: The full discussion provides additional context on how machine learning is applied in energy markets, including detailed examples from real datasets and research projects. It also expands on the practical challenges of working with energy data across Europe. Watching the full conversation can be useful for professionals who want a deeper understanding of how modeling assumptions and data limitations influence decision-making in energy systems.

From Data to Energy Systems: A Practical Entry Point

Chiara’s path into the energy sector was driven by a clear objective: using data to support the transition to more sustainable energy systems. Her background in applied mathematics and early experience working with energy utilities exposed her to the scale and complexity of electricity markets.

Consulting work provided practical exposure to project development, including how renewable assets are evaluated and financed. However, it also revealed a limitation common in commercial environments: limited time for deeper analytical work. This led her to pursue academic research focused on building more robust analytical methods for understanding energy systems.

Her current work centers on analyzing large-scale operational data to understand how power plants behave in real market conditions.

What Operational Data Reveals About Market Reality

A key focus of Chiara’s research is the use of time-series data from power plants to better understand system behavior. In one study, she analyzed five years of hourly data from gas power plants in Germany to assess their operational flexibility.

The expectation in a system with high renewable penetration is that conventional plants operate as flexible backup, generating electricity only when the renewable supply is insufficient. However, the data revealed a different picture.

Approximately half of the analyzed plants were not operating as flexibly as expected. In some cases, they continued generating electricity even during periods of low demand or negative prices.

This highlights a gap between how models assume assets behave and how they actually operate in practice.

Hidden Constraints in Energy Modeling

Traditional energy-market models often focus on technical capabilities, such as ramp rates or efficiency. However, Chiara’s work points to a broader set of constraints that are rarely captured. These include:

• Maintenance agreements that discourage flexible operation

• Contractual obligations with local utilities

• Operational practices tied to asset longevity

• Strategic or commercial considerations

These factors are not typically visible in datasets and are difficult to formalize in models. As a result, models may overestimate system flexibility and underestimate the persistence of conventional generation.

For practitioners, this creates a clear risk: relying on models that do not reflect real-world behavior can lead to incorrect assumptions about system dynamics and investment decisions.

Market Power as a Structural Issue

Another area of Chiara’s work focuses on market power in electricity markets. Market power arises when large generators can influence prices, particularly during periods of supply scarcity. Her research shows that:

• Market power can be detected using empirical data

• It can be simulated using advanced methods such as reinforcement learning

• It can be mitigated through regulatory design

Comparisons between U.S. and European markets illustrate the role of regulatory frameworks. In the U.S., structured mechanisms such as ex-ante mitigation procedures are used to limit excessive pricing. In contrast, Europe remains more fragmented, with national-level approaches still dominant.

This fragmentation makes it harder to implement consistent safeguards across markets.

For companies operating in multiple regions, this means that pricing behavior and regulatory exposure can vary significantly depending on location.

Data Availability Is Not the Main Problem

Contrary to a common assumption, Chiara emphasized that the energy sector does not lack data. In fact, there is extensive availability of open data across European and U.S. markets. The main issue lies elsewhere: the lack of standardization. Common problems include:

• Inconsistent naming conventions for assets

• Multiple identifiers for the same power plant

• Missing or incomplete metadata

• Human errors in data registration

These issues make it difficult to merge datasets and build consistent analytical frameworks. The problem becomes more severe as data becomes more granular, for example, with distributed solar assets.

For organizations, this reinforces the importance of investing in data cleaning, mapping, and validation processes rather than assuming raw data can be used directly.

Why Model Outputs Should Be Treated Carefully

Chiara highlighted two fundamental limitations that practitioners should keep in mind when working with models:

1. Models depend on assumptions
   Every model is built on simplifying assumptions about behavior, correlations, and system dynamics. If these assumptions do not reflect reality, the outputs will be misleading.

2. Correlation does not imply causation
   Statistical relationships identified by models do not necessarily reflect causal mechanisms. Misinterpreting these relationships can lead to incorrect conclusions.

This is particularly relevant in energy markets, where multiple variables interact in complex ways across time and geography.

For decision makers, the key takeaway is that model outputs should be interpreted as scenarios or approximations, not definitive predictions.

Where Machine Learning Adds Value and Where It Does Not

Machine learning is increasingly used in energy systems, but Chiara stressed that its role should be clearly defined.

Areas where machine learning adds value:

• Forecasting, especially for prices and demand

• Pattern recognition in large datasets

• Simulation and control applications, such as optimizing EV charging

Areas where traditional methods remain dominant:

• Market clearing and dispatch optimization

• Long-term capacity planning

• Grid-constrained optimization problems

These areas are fundamentally numerical optimization problems, where established methods remain more reliable.

Machine learning can still support these processes, for example, by providing initial estimates or improving computational efficiency, but it does not replace core optimization frameworks.

This distinction is important for companies evaluating where to invest in advanced analytics capabilities.

Market Design Requires Less Dogma and More Evidence

From a market-design perspective, Chiara emphasized the need for a more flexible and evidence-based approach.

Debates around energy systems often become polarized around specific technologies, such as gas versus nuclear. However, different countries face different constraints, and no single solution applies universally. Instead, effective market design should be guided by:

• Data and empirical evidence

• System-specific constraints

• Evolving technological capabilities

• Practical feasibility

This approach is particularly relevant in Europe, where energy systems are interconnected but still governed by national-level decisions.

The Role of Interconnectivity and System Effects

One underexplored area is the impact of cross-border transmission and interconnectivity. While interconnection can support renewable integration by balancing supply across regions, it can also create unintended consequences. For example, countries with historically lower prices may experience price increases due to integration with higher-cost markets.

This can affect public acceptance of the energy transition, especially in regions that were already relatively low-carbon.

For policymakers and companies alike, this highlights the importance of considering system-wide effects rather than focusing only on local outcomes.

Skills Needed for the Next Generation of Professionals

Looking ahead, Chiara pointed to the need for broader, interdisciplinary skill sets. The energy sector has traditionally been divided between engineering and economics. However, future challenges require professionals who can bridge these domains and understand both technical systems and market dynamics. Key capabilities include:

• Data analysis and modeling

• Understanding of market design and regulation

• Knowledge of physical energy systems

• Ability to work across disciplines

This shift reflects the increasing complexity of energy systems and the need for integrated approaches.

Conclusion

Chiara Fusar Bassini’s insights highlight a central theme: the gap between models and reality remains one of the key challenges in the energy transition.

Operational data shows that assets do not always behave as expected. Data is available, but often difficult to use effectively. Machine learning offers clear benefits, but only in specific contexts. Market design remains fragmented, and system-wide effects are still not fully understood.

For companies and decision makers, the practical implications are clear:

• Do not rely on models without understanding their assumptions

• Treat data quality and standardization as core priorities

• Apply machine learning selectively, not universally

• Consider system-level effects when evaluating market developments

As energy systems continue to evolve, combining data, domain knowledge, and realistic assumptions will be essential for making informed decisions.