pes-summerschool@dtu.dk

Program

Keynotes and lectures

Poster session

Panel session with experts

Social activities and networking

Lunch and snacks

Keynotes and lectures

During the 6-day course distinguished speakers, excelling both in research and teaching, give talks about the latest developments in research for energy systems. By clicking in the buttons below you can find the details of each of the summer school’s talks and their presentation.

Styled Buttons
Title: 50 Years of Market Design Theory & Experience in the US: The Challenge of Supporting Investment with Spot Prices

Abstract: Since Schweppe’s famous “Power Systems 2000” article (IEEE Spectrum, 1978), the fundamental principle that convex optimization problems yield prices that are revenue adequate and support optimal decisions has been a “lode star” for guiding market designs.  A half century of trying to apply this principle when real markets are characterized by lumpiness, scale economies, market power, imperfect information, price-distorting regulation, externalities, and other market failures has yielded a number of lessons learned in US markets.  Changes in technology threaten to outpace how fast we learn these needed lessons.   I will reflect on how far US markets have come, and remaining challenges of realizing all the potential benefits of variable renewables, storage, and demand response.  These reflections are biased by my 24 years of experience on the Market Surveillance Committee of the CAISO trying to coax prices from market software and processes in order to support optimal operations and investment in the western US markets.
Title: Congestion Management in Electricity Markets

Abstract: The lecture introduces price formation in liberalized electricity markets, using the Nordic and European power markets as its main reference point. It covers marginal cost pricing under transmission constraints, the fundamentals of congestion management, nodal versus zonal market design, including flow-based market coupling, and illustrates the methods by means of stylized network examples. It discusses experience with these approaches to date, and current challenges and possible ways forward.
Title: Long-Duration Energy Storage: Investment Challenges and Hedging Contracts

Abstract: Long-duration energy storage (LDES) is increasingly recognised as a critical technology for future deeply decarbonised power systems. However, uncertainty and variability in investment returns can discourage investment in LDES . To address this challenge, policymakers are introducing revenue-hedging mechanisms aimed at reducing investment risk. In this talk, Elina will discuss how risk-averse stochastic equilibrium models can be used to quantify barriers to LDES investment and evaluate the effectiveness of revenue-hedging contracts. Drawing on a case study of Great Britain, the talk will illustrate how mechanisms resembling GB’s cap-and-floor scheme influence investment decisions, system-wide outcomes, and the distribution of costs and benefits among market participants. The talk will conclude by discussing potential avenues for future research.
Title: Algorithmic collusion in short-term trading, learning dynamics, and market outcomes

Abstract: Strategic decision-making, such as optimal pricing, is increasingly delegated to machine learning algorithms. This lecture investigates how autonomous learning algorithms can generate and sustain supracompetitive market outcomes without explicit communication. Drawing on an application to the short-term electricity trading of batteries, we show that autonomous reinforcement learning (RL) algorithms can operate batteries in ways that maximize their joint profits at the expense of consumer surplus, without any communication. We demonstrate that this outcome, which resembles collusion, is persistent and investigate possible ways to mitigate it. We then develop a theory of algorithmic cooperation with Q-learning to examine the relationship between the exploration strategies of RL algorithms and their propensity to cooperate, and relate our findings to existing results in the literature. We also investigate how different aspects of algorithmic sophistication affect cooperation. Finally, we study the outcomes of algorithms learning under conjectural variations, i.e., when they estimate competitors’ reaction functions while optimizing their own prices. We theoretically show that the sequential nature of learning strongly affects the frequency of supracompetitive prices and present a series of simulations with large language models (LLMs) to support our theoretical findings. We conclude by discussing the implications of these results for competition policy and proposing potential regulatory responses.
Title: Algorithmic Trading and Multi-Market Strategies for Battery Storage in Short-Term Electricity Markets

Abstract: The transition towards renewable energy has fundamentally changed short-term electricity markets. Increasing shares of intermittent wind and photovoltaic generation have amplified price volatility and created a growing need for flexibility. Battery Energy Storage Systems (BESS) are particularly well suited to provide this flexibility. However, operating batteries profitably requires sophisticated decision-making in increasingly complex and fast-moving electricity markets. This lecture provides an overview of algorithmic trading and multi-market bidding strategies for energy storage. We will begin by reviewing the fundamentals of modern short-term electricity markets and the current state of grid-scale battery storage, including capacity developments, market participation opportunities, and technical constraints such as battery degradation. Building on these foundations, we will explore the design and implementation of algorithmic trading strategies across day-ahead auctions, continuous intraday markets, and ancillary service markets. Particular attention will be paid to the trade-offs between computationally efficient heuristics and more sophisticated optimization approaches, the role of real-time decision-making, and the integration of price and weather forecasts. Finally, we will examine the challenges of coordinating battery participation across multiple markets. We will discuss how decisions in day-ahead auctions, continuous intraday trading, and ancillary service markets interact, and how the resulting operational and economic trade-offs can be addressed through integrated bidding and trading strategies. The lecture aims to provide PhD students with both the theoretical foundations and practical algorithmic insights needed to understand and research the evolving role of energy storage in electricity markets.
Title: Data Privacy and Fairness (tentative)

Abstract: TBD
Title: The European electricity market model: achievements, emerging challenges and regulatory responses

Abstract: Over the past two decades, Europe has made significant progress towards an integrated electricity market, built around cross-border market coupling, marginal pricing and increasingly coordinated balancing arrangements. Yet, recurrent external shocks, renewed episodes of high electricity prices and growing concerns about security of supply have brought fundamental questions about the effectiveness of the European electricity market model back to the forefront of the policy debate. While the recent electricity market design reform has largely preserved the architecture of short-term markets, it has also strengthened the role of long-term contracts and capacity remuneration mechanisms. This raises a question about whether the current framework can provide efficient medium- and long-term price signals, support adequate investment in generation and flexibility, and preserve market integration beyond short-term trading. There is a risk that increasingly national approaches to investment support and resource adequacy could lead to greater fragmentation of the European electricity market over longer time horizons. Drawing on European regulatory experience, I will discuss some of the key challenges ahead: the interaction between short-term markets, long-term contracting and capacity mechanisms; the growing importance of flexibility and how the market power mitigation and enforcement interact with market design choices.
Title: Trustworthy Foundation Models for Power Grids

Abstract: AI Foundation models have the potential to drastically accelerate fundamental functions for power grids, such as power flow, state estimation, optimal power flow and power system dynamics. However, to find application in real power grids, they need to be trustworthy. This talk introduces methods and discusses directions that equip power grid foundation models with the necessary trust. These include statistical guarantees, physics-based guarantees through formal verification methods, and additional layers which can boost the reliable performance of foundation models (and LLM agents) through traceability and auditability. Our goal is to remove the barriers for the large scale deployment of safe AI in power systems.
Title: Learning on Shared and Uncertain Data: Privacy-Preserving and Resilient Methods for Forecasting and Market Participation

Abstract: Sharing data among energy system actors offers significant opportunities to improve forecasting accuracy, operational efficiency, and electricity market participation. For example, sharing spatio-temporal measurements across renewable energy sites can improve forecasts and reduce the associated costs for market participants and grid operators. However, exchanging raw data is often restricted by privacy, confidentiality, and commercial constraints. Federated learning (FL) provides a promising framework for collaborative model training while keeping data locally with their owners. Nevertheless, conventional FL architectures do not inherently guarantee privacy, as sensitive information may still be inferred from shared model parameters or updates. The first part of this lecture will introduce privacy-preserving collaborative learning methods for energy forecasting and electricity market participation, covering both encryption-based and encryption-free approaches. The second part will address resilience to imperfect operational data, exploring methods to mitigate the impact of missing, corrupted, or delayed measurements on predictive performance in multi-actor and multi-asset energy systems. The lecture will conclude with emerging research directions for learning from distributed, confidential, and imperfect data, including the potential of foundation models for energy forecasting and other power system applications.
Title: PPA Design: Bargaining over Contracts and Sharing Imbalance Costs

Abstract: Power purchase agreements (PPAs) make renewable projects bankable, but their terms must be negotiated between parties with different risk preferences and asymmetric beliefs about uncertain parameters such as future market prices. We first review common PPA designs and show how Nash bargaining can jointly determine the strike price and contracted volume. We then take the perspective of a facilitator that brokers PPAs between renewable producers and consumers and manages their residual imbalances, and formulate the sharing of the resulting imbalance cost savings as a cooperative netting game.
Title: Information is in Control

Abstract: Traditionally, decision making begins with people gathering information and choosing how to act. Digital platforms are fundamentally changing this process by determining what information people see and using it to shape their decisions. From navigation and online shopping to social interactions and markets, AI is dramatically expanding platforms’ ability to influence the behavior of millions of people and autonomous agents. To influence decisions effectively, platforms must understand users’ private preferences, objectives, and constraints. While these can be learned from data, effective personalization often requires eliciting private information in real time. This gives rise to a new control paradigm based on information markets, where platforms design personalized information to coordinate autonomous agents toward collective objectives. Such markets require mechanisms for eliciting private information, aligning incentives, and learning, with feedback loops involving both information and payments. This paradigm presents both promise and peril. Personalized information can improve decisions and system performance, but it can also enable platforms to exploit information asymmetries, manipulate behavior, and exercise market power. This talk will explore how information can be designed to influence collective outcomes, when such design serves societal objectives, and when it becomes a mechanism for manipulation. Understanding this duality is essential to realizing the benefits of AI-enabled platforms while limiting the risks of concentrated informational power.

The full schedule of the summer school can be found below:

Poster session

Please prepare a poster on your current, past or future research! The two best posters selected by all participants on Sunday and Monday will have a chance to be presented orally on Thursday or Friday. Information on poster printing and presentation schedule will be shared closer to start of the summer school.

Social activities and networking

Multiple events during the summer school are dedicated to giving you the opportunity to see what Denmark has to offer as well as allow you to meet your fellow summer school participants and foster relationships that can build your network and hopefully lead to future research collaborations.

Lunch and snacks

To make sure you have enough energy throughout the day we provide vegetarian lunch and snacks for all participants of the summer school.