长短相形，高下相倾。

Long and short take shape by comparison;
high and low depend on each other.

《道德经》第二章

How many day-ahead spread multiples is flexibility worth?

Read more

On 6 April 2026, Easter Monday, Belgian imbalance prices hit the regulatory floor of -15,000 EUR/MWh for five consecutive quarter-hours, from 13:45 to 15:00 CEST.

Let’s try to trace the chain as far as public data allow: Belgian imbalance settlement, the mFRR scheduled-down component, MARI, German downward mFRR, and the possible French/RTE balancing layer that ordinary border flows do not show.

Then it’s worth reflection on the economic rationale for extremely negative bids. If a downward mFRR bid was accepted at -15,000 EUR/MWh, what kind of asset constraint, portfolio exposure or market-design feature could justify it?

Read more

Start with the domestic generation mix over the last year. Czech nuclear averages about 3.4 GW and lignite about 2.6 GW, against average load around 7.2 GW. Solar matters, but the backbone is still nuclear plus lignite.

Czech and German day-ahead prices sit close to each other, as you’d expect. In May to date, the daily DA average is about EUR 98/MWh in Czechia and inGermany.

Czechia and Germany are, of course, both cleared inside Single Day-Ahead Coupling. EUPHEMIA takes orders from the power exchanges and network constraints from TSOs, matches supply and demand, allocates cross-zonal capacity implicitly, and returns prices, trades, net positions and scheduled exchanges.

Prices converge when the market can move energy across the relevant borders. They diverge when cross-zonal capacity or flow-based constraints bind, when local bid curves are very different, or in operational fallback and decoupling situations.

Read more

Rank 2 of 844 by average price in our sample. Rank 3 by marginal accepted price.

Read more

I wrote last time how this keeps re-occurring to me as a simple motto.

Here’s a consequence:

A long post does not only ask for attention. It asks for a block of attention.

If you can’t find that block, the post becomes a deferred task.

Deferred tasks become backlog.
Backlog becomes guilt or avoidance.
Avoidance makes the subscription feel less valuable.

So we want shorter formats to complement the longer posts.

Heat is in the European news.

Western Europe is moving into the first real warm spell of the summer season. France, Spain and Portugal are the louder story; Germany is not the epicentre.

That makes Germany more interesting, not less.

We are not looking at an emergency weather setup. We are looking at how a warm, bright weekend translates into market shape.

Warm, bright, dry, low wind: open-Meteo has maximum temperatures up to about 30.6 degrees C on Saturday and 31.3 degrees C on Sunday across a sample of German stations.

Shortwave radiation averages about 26.8 kWh/m2 on Saturday and 26.2 kWh/m2 on Sunday. Cloud cover is low on Saturday, about 16% on average.

Read more

Introductory Reflection

The New York Times yesterday had a great piece on Dwarkesh Patel and his podcast. He is a well-connected AI-optimist, each episode is deeply researched, each a starting point for further reference and exploration. Material like this is vital, I think, to help us reflect on the rapid changes.

I worry sometimes about my thinking habits in the age of AI. Where is the boundary between the work that the AI can help with and a sense of ‘artful pursuit’ that provides orientation, meaning, a golden-thread?

The main challenge, I believe, is to train and maintain attention. “Attention is all you need” keeps re-occurring to me as a motto, suitably appropriated out of its original context, as I find myself snapping out of spells of distraction.

Our attention here is focused on exploring the energy industry with the new technology, to build understanding, connection, conversation with increasing momentum. My aim is to share a combination of energy-related content (“the what”) combined with thoughts on the machinery used to develop it (“the how”). The weekly market updates are a litmus test of progress.

Today’s market update is once again Germany focused but includes cross-border analysis. International exploration will develop step-by-step… The content was generated entirely by the ‘Stratnergy Exploration Engine’ but the process involves a number of ‘tuning’ and ‘editing’ passes. (One of the hardest problems for AI is to figure out what problem to solve, and which line of enquiry to pursue.) Improvements in these updates require a significant amount of work, choice, and an increasing budget threshold for token ‘intelligence.’

“The idea becomes a machine that makes the art.” — Sol LeWitt

Stratnergy Market Update

No extreme price event this week. That is the starting point. DA, ID-AEP and SPARX all cooled after the early-May drama, but 15 and 19 May still brought sizeable downward balancing calls. We start with context (the familiar boxplots), then the forecast misses, the aFRR- TSO footprint, frequency, and finally the border clock.

From 13 to 19 May, the day-ahead average was about EUR 106.7/MWh. The average daily high-low range was about EUR 151.6/MWh, down from about EUR 210.8/MWh over 1-12 May. The lowest quarter-hour in the recent window was only EUR -2.33/MWh.

The longer history shows how quiet the price layer became after early May.

ID-AEP and SPARX also calmed. Recent SPARX averaged about EUR 214.2/MWh, versus about EUR 363.4/MWh over 1-12 May. The largest recent SPARX was EUR 360.5/MWh.

Capacity prices came down from the early-May peaks. FCR 12-16 averaged about EUR 160.3/MW/4h, down from about EUR 250.0/MW/4h over 1-12 May. aFRR- 12-16 averaged about EUR 243.1/MW/4h equivalent.

aFRR- activation was high on 2026-05-15: about 2,995 MWh; 2026-05-19: about 2,966 MWh. These were not extreme-price days. The system can still need significant downward correction when the visible price readout looks ordinary.

Since the start of 2025, 15 May ranks 33 of 504 operational aFRR- days by total activation, and 19 May ranks 34. The pair separates two things that are often blended together: balancing activity and realised price stress.

Read more

Memes wrap around grains of truths. Stories spread and mobilise capital, they get entrepreneurs and builders motivated to focus around the next big thing. Momentum needs simplification. It’s how we’re wired.

But simplification becomes problematic as it hardens and as we lose sight and understanding of what is actually going on. Shadow-play in a cave: the customer disappears behind click-thru rate statistics. The borrower is anonymous behind the tranche. The data centre seems unreal, covered in conceptual clouds. The abstraction begins to take the place of the thing it describes; the derivative overshadows the underlying. When’s the last time you thought about Poynting vectors?

Carlota Perez’s work on technological revolutions describes moments when finance and technology reinforce each other: money arrives, expectations overshoot, infrastructure gets built, and it’s only later, after dust settles, that we understand which parts of the new system are here to stay. The overshoot is not simply a mistake. Sometimes it is how a new system gets installed.

Batteries are necessarily somewhere in a pattern of this kind. That’s worth thinking about if you are in any way interested in the future of the energy system.

For a while, grid-scale storage looked like one of the cleanest stories in the energy transition. Solar build-out would deepen the daily price shape. Midday power would become cheap, evening power expensive. The shape of the profile looks a bit like a duck. A perfect meme. Batteries would follow the duck curve: charge when prices were low, discharge when prices were high, add some ancillary revenue. Volatility as a business model, guided by software, by AI, optimising and trading in free markets with the latest technology. The purest possible play on the energy transition.

I’m not being cynical, quite the opposite.

The story was attractive because it was simple, and because it contained truths: renewable generation does change the shape of power markets, volatility does create value for assets that can move energy through time, batteries are one of the few technologies that can respond quickly enough to matter.

Reality has a way of catching up to the first-order abstractions, however. Network location, market rules, local constraints, forecasts, the behaviour of other assets responding to the same signals. The value is in the relationship between the asset and the system. The play is not as clean as many thought. A smell of disappointment is in the air, sometimes. Have you noticed?

What is the way forward? Openness. Not openness as branding or as insider networking, nor as a fantasy that everything should be free and accessible, but openness as a way for complex mechanisms to become exposed for interaction. More participants. More inspection. More shared definitions. More ways for the grid, the asset owner, the operator, the regulator, the customer, and the market to correct one another.

Abstraction works when reality can loop back in to move the concept forward.

Read more

Solar pressure, wet/cloud softness, wind and forecast risk, border flows, and the evening handover after PV fades… the weather moves us through a range of different patterns. Let’s count. Since the start of May we’ve had 2 clear/high-radiation days, 5 wet/cloud days, 2 wind/cool days and 4 mixed days.

To clarify: Clear/high-radiation means shortwave radiation above 20 kWh/m2 with limited cloud and little rain. Wet/cloud means either material rain or very high cloud with low radiation. Wind/cool means maximum wind speed above 22 km/h with average temperature below 10 degrees C. Entirely arbitrary but useful distinctions.

What were things like this time last year? Skip this if you’re not following the markets daily… May 2026 has a DA average around EUR 97/MWh compared with EUR 68/MWh over the same first thirteen May days in 2025. The average DA high-low range is EUR 204/MWh, versus EUR 179/MWh last year. FCR 12-16 is averaging EUR 244/MW/4h, above both the same 2025 window at EUR 213/MW/4h and the 2026 year-to-date average of EUR 118/MW/4h. aFRR- 12-16 is about EUR 327/MW/4h equivalent, close to the same May window in 2025 but well above the 2026 year-to-date average of EUR 138/MW/4h equivalent.

The May detail shows a low-price spring shape that keeps changing character. Clear days do not automatically become negative price risk events. Wet/cloud days can calm the DA lower tail while reserve capacity remains firm. Wind/cool days move the focus toward residual load and forecast error.

On the realised intraday side, May 2026 is not far from last May on the average readout: SPARX is running around EUR 363/MWh for the complete days through 12 May, versus EUR 367/MWh in the same 2025 window. The ID-AEP daily range is similar too, around EUR 444/MWh versus EUR 455/MWh. The difference is less the average realised range than the way reserve prices remain elevated while the weather keeps moving between regimes.

Thinking about forecasts

"Lovely day, isn't it?", "Can you believe all this rain?", "How about this weather?" Polite, innocuous conversation, you might think. Nein! The weather is a serious matter, particularly in Germany: a ‘lovely’ mix of shortwave radiation and cloud cover translates to complex PV expectations; wind becomes generation and forecast-error risk; temperature becomes load; rain and river conditions become hydro, cooling and availability questions. Those expectations then interact with residual load, grid constraints, reserve procurement, cross-border capacity and trader positioning. Prices trace these interactions as do flows, activations and frequency deviations.

Shortwave radiation has the relationship with midday DA price that you’d expect. Reserve prices also move with the weather, but not as the same signal: the scatter and the exceptions are where it gets interesting

Open-Meteo deserves a bow here. Patrick Zippenfenig’s project takes open numerical-weather data from services such as DWD, NOAA, Meteo-France, ECMWF and others, processes more than 2 TB of model data per day, and exposes it through a no-key JSON API.

Open-Meteo

The latest developments around weather forecasts!

By open-meteo

Please show your support by signing up to the open-meteo Substack, which is relatively inactive at the moment, unfortunately, but who knows what the power of new subscriptions might lead to! The source code is public. Open infrastructure in the best sense: quiet, useful, inspectable. Kudos, !

The recent run of the weather and the corresponding energy situation is mixed and rich. The system has several routes into tension. Three hypotheses you might like to have in mind:

Read more

Who are the trading desks in this emerging market, and what do they do?

Our starting point is a LinkedIn conversation in which Fortinbras and FlexPower were described as trading desks responding to volatility with new instruments: day-ahead swaps, BESS revenue floors, short-term PPAs.

FlexPower and Fortinbras are recognisably ‘desk-like’ in a financial sense: they structure risk, create products and may take positions.

But what about Entrix, ESFORIN, enspired, suena, or a virtualiser such as terralayr? They also perform trading-related functions. Are they ‘trading desks’?

Perhaps we need to expand our idea of what a trading desk can be: technology-led optimisers, aggregators, and virtualisers may soon be at the forefront of creating a new kind of trading desk. The power trading desks of the future might focus on developing ways for software-agents to originate, structure, execute, and manage variability swaps based on objective functions and remits set by their principals.

But before we become even more imaginative, let’s try to tie down the notion of a ‘trading desk’ in the energy context through analogy to finance.

In finance, a flow desk handles client trading. A prop desk trades the firm’s own book. A structuring desk designs products. A quant desk prices them. A risk desk constrains them. Index providers define a reference values. There are trading venues and clearing houses. Let’s think about energy industry analogues.

A desk is a function, not a company

A desk is a commercial function: a place where a certain kind of decision is made, price is formed, risk is owned, contracts structured, flow is executed, or exposure is transformed.

Energy firms often bundle several desk functions together. One company may optimise the asset, support structuring, provide the trading algorithm, advise on bankability, and perhaps even take some market risk. Conversely, one contract may split these functions across several parties: asset owner, optimiser, offtaker, lender, benchmark provider, and trading venue.

A desk answers a question.

Who trades the flow?
Who owns the volatility?
Who creates the floating revenue?
Who writes the product?
Who defines the reference value?
Who hosts the trade?
Who stands behind the risk?
Who controls the physical asset?

As the markets mature we may see clearer lines around these questions and corresponding specialisation.

Energy-desk taxonomy

The European flexibility market is not producing one new kind of firm. Instead, it is beginning to develop a ‘stack’ of functions.

It starts with the asset desk. This is the owner or developer of the physical exposure: batteries, renewable portfolios, CHP, industrial loads, distributed batteries, flexible demand, co-located assets.

Then comes the optimisation desk where physical flexibility connects to market dynamics: when to charge, discharge, reserve, rebid, hold back, or stay idle. Trading algorithms need to work against state of charge, degradation, warranty, grid connection, market access, imbalance exposure, and operational constraints.

The asset and optimisation desks set the framework for the atomic question: to cycle or not to cycle? Every decision sits at the intersection of physics, economics, and governance: is the asset within limits, is the spread worth capturing, and is the action permitted?

Next comes the flow trading desk. This is where short-term power is bought and sold. Day-ahead, intraday, balancing, ancillary, shape, spread, imbalance. The flow desk is where the asset’s optionality becomes market action.

Read more

Labour Day was a test-case for one kind of stress scenario: low demand, strong solar, a deep day-ahead lower tail, expensive downward reserve readiness, and ID-AEP/SPARX dislocation from the expected story.

Read more

We move from preview to a review, which reminds us that ‘whatever has the nature to arise, will pass away.’ Eventually, certainly. If you preview, you must review. “Always follow up,” I tell myself.

The labour movement knew the power of follow-up, but its point was to win the right to stop. There’s some danger in my little maxim: every loose end can become a summons, every unanswered email a moral debt, every task a tiny foreman. So, follow up enough, but let some things pass. Work needs diligence, but life needs some drift.

What we’re doing here isn’t work, however. Work is what you do for others, art is what you do for yourself as Franz points out in Sondheim’s play on Seurat. Following the power markets together is a pursuit, a challenge, a drifting conversation, supported by the new systems we now have to help us explore.

Our use of these systems to explore the energy context can be seen as an exploration of a fundamental question: how can or should we calibrate the exploration of our interests, be they at work or in ‘art’ or leisure, to the shockingly quick development of the new technology?

“I think for the actual mental load of being a person, interacting with these systems is going to be quite strange. I already find that I’m like: Am I keeping up with the ability of these systems to produce insights for me? How do I structure my life so I can take advantage of it?” Jack Clark

When in doubt, we must always come back to reflect on whether we are having ‘fun,’ the definition of which is, of course, highly personal.

“It’s hard to compete against someone who’s just there to have fun.” Peter Steinberger

Now that we’re wound up, let’s review Labour Day in full. It’s such a useful and important example, calibrating us to the season and the times ahead.

• The generation profile, imports/exports;

• DA expectations vs ID actuals;

• Ancillary activations;

• The value of variability swaps if May-day extremes continue.

For context and to explain why the example is so important, consider the discourse around the energy situation we find ourselves in. It’s worth tracking the full range of arguments and perspectives: cheap renewables, missing flexibility, market coupling, consumer opportunity, grid stress, political point-scoring, various agendas.

The situation is highly charged. Some perspectives are well developed, some seem too quick. In either case, it’s more important than ever to develop the conversation responsibly, productively.

• “Symptom von zu wenig Flexibilität im Stromnetz und zu wenig Speicherkapazitäten,” Windkraft-Journal

• “schaltet morgen eure PV-Anlage ab,” Lion Hirth on LinkedIn; “vom Lastfluss über Spannung bis hin zur Systembalance,” NIUS, quoting Lion Hirth; “mehr Strom [im] Netz… als wir abnehmen können,” Lion Hirth on LinkedIn

• “Duitse producenten gaan dan proberen om hun stroom bij ons te slijten,” ANWB Energie

• “Price drops are not limited to Belgium,” The Brussels Times / Belga

• “více elektřiny, než se stihlo spotřebovat,” FXstreet.cz, reporting Bloomberg

• “Deutschland Geld ins Ausland zahlt,” AfD Bundestag press release. For a very interesting wider debate see also: Philipp Man’s challenge to Daniel Stelter.

• “Risiko eines Brownouts,” Apollo News or “was die Energiewende bringen kann und wo es hakt,” Golem.de, etc. etc. etc.

Hirth’s intervention is narrower than many of the more politicised perspectives. The argument is not that solar is the problem. It is that a large volume of small, weakly controllable PV is unhelpful when demand is low, prices are deeply negative, and system margins are already compressed. That is a flexibility and controllability problem, not a reason to flatten the whole debate into a verdict on renewables.

So, given this starting point, consider the fundamentals. (Please excuse the gaps in the load data, which are from the source. You may have noticed the same on Energy Charts.)

Solar peaked at 56.6 GW around 11:15 Berlin time. Net exports peaked at 9.0 GW around 12:30; by the evening the line had flipped to net imports, peaking at 6.9 GW around 20:00. This helps explain why, as we will see later, aFRR- activation was fairly small: the midday surplus was expected and fed through into negative prices and net exports, rather than as a large national downward-balancing activation. The activation event came later and in the other direction.

The 12-16 export block went mainly towards Denmark, Austria, Norway and the eastern borders. At the 12:30 export peak the leading destinations for German flow were Denmark 3.1 GW, Austria 1.5 GW, Norway 1.4 GW, Czech Republic 1.0 GW. Over the full delivery day, the net export balance was led by Poland 28 GWh, Netherlands 27 GWh, Denmark 20 GWh, Czech Republic 18 GWh, while Germany was a net importer from France 47 GWh, Switzerland 32 GWh.

It’s important to monitor these cross-border flows, because power transmission creates political linkage too. Remember Norway? In Norway to Texas via Yunnan, I., we saw how cross-border electricity integration can turn someone else’s energy transition into your domestic politics. That has not gone away.

• “Use May 1st to make clear your opposition to the EU’s overriding of Norwegian energy policy. Demand national control over power. Norway needs industry!” Nei til EU Vestfold; “utenlandskabler til Tyskland og Storbritannia ... import av europeiske strømpriser,” Nei til EU Vestfold

• “no to new electricity interconnectors,” regjeringen.no; “does not want Norway to align itself more closely with the European energy system,” regjeringen.no

• “Strømprisene har vært usedvanlig høye i første kvartal,” Teknisk Ukeblad / NTB, citing NVE

• “redusert etterspørselsrespons,” Finansavisen / DNV authors

In some places at least, Norwegian arguments about interconnectors hardened into Labour Day language about national control, industry and imported European prices. While our pivot point here is Germany, we must develop the wider European context. Physical flow feeds into political flow.

So, moving back to our central pivot, let’s focus in further on the day. How did the fundamentals play out in the markets? How can we take Labour Day as a prompt to think about the ‘job’ of flex and the expected values of variability swaps?

Read more

The day-ahead average is EUR -2.08/MWh, but that’s probably the least interesting number here. The DA low is EUR -499.99/MWh, the high is EUR 232.99/MWh, for a high-low spread at EUR 732.98/MWh. The 12-16…

Read more

This post will funnel us from an overview of the year so far to a view of the current week in context, and we’ll end by looking forward. The familiar box-plot will be useful as ever: it tells us whether the middle moved, whether the lower tail opened, whether the upper tail stayed alive, and whether the current situation is sitting inside a familiar distribution or outside it.

We want to look at the whole market banquet, not just the individual dishes.

To support all of this we take advantage of data-source and pipeline developments over the last weeks. To highlight one example yet again, we had run into a wall with ID related measures like SPARX because the daily overview data published by EPEX is not suited for regular use.

The re-discovery of ID-AEP has opened up all options again. ID-AEP gives us a perfectly suited realised-price trace of the continuous ID market.

Before zooming into 2026, it helps to step back into 2025.

Read more

Well, you tell me whether this is useful. It's certainly not easy to generate. The content is the result of multiple iterations adding up to significant overall run-time. Editing remains very important. Graphics are becoming better, but there’s a point at which I stop optimising, as you’ll see below. The language mirrors some of the language used in Wardley/ Value-Chain-Mapping, e.g. ‘games.’

The short thesis…

The German flexibility market is differentiating into several control-point ‘games.’ Optimisers fight for the dispatch decision. Virtualisers fight for the asset-access interface. Structurers fight for the risk book and benchmark. Asset owners fight for sovereignty, lifetime value and bankability. Utilities and BRPs fight for hedge, balancing and customer-control positions. DSOs and TSOs pull everything back toward the node, connection, tariff and system-security constraints.

The battery is physically the same asset in each story. Strategically, it is not the same component. To a trader it is optionality. To a virtualiser it is capacity. To a financier it is a portfolio. To a DSO it is a node. To a retailer it may become a tariff proposition. To an open-source strategist it is a bundle of interfaces, data standards and testable operating claims.

That is why the value-chain-mapping lens helps. It allows us to integrate the perspectives of platform, aggregator, optimiser, trader through a unifying set of questions: which user need is being served, which value-chain component is being controlled, and how is that component evolving?

Why revisit previous episodes?

The earlier Compare & Contrast posts did useful work. Part I separated Suena, Entrix, Enspired and ESFORIN as different optimisation and trading offers.

Part II added terralayr and FlexPower, which forced a sharper distinction between virtualisation and structuring.

The later flexibility updates added capital flows, global trading houses, integrated developers, Kyon / TotalEnergies, Terra One, newer or potential new entrants like Tesla and Habitat, and the possibility of new performance-transparency norms.

The platform-thinking series gave us theoretical scaffolding: pipelines, platforms, aggregators, inverted firms, modularity, open interfaces and the energy smiling curve.

This revisit changes the unit of analysis. It asks where each actor sits on the value chain, what control point it defends, and which parts of the stack should be industrialised by open standards or open-source reference implementations.

The Value Chain (Wardley) Reset

A Wardley map has two disciplines that are especially helpful in energy.

The first is value-chain discipline. Start with the user need. Work downward into the capabilities and components required to meet that need. Visibility falls as we move down the chain: investor confidence is more visible than telemetry protocols; revenue stack is more visible than a BMS API; tariff design is less visible to a trader than to a DSO, but it may still dominate the outcome.

The second is evolution discipline. Components move from genesis to custom-built to product/rental to commodity/utility. That movement changes management method. Novel things need exploration. Custom things need learning and adaptation. Productised things need vendor selection, contracting and performance comparison. Commodity things need cost control, resilience and operational excellence.

The strategic questions then become straightforward but uncomfortable.

• Why are we treating a commodity component as if it were proprietary?

• Why are we outsourcing a novel component before we understand the need?

• Which component is moving rightward faster than our business model assumes?

• Where are we confusing a product, a service, an aggregator and a platform?

• Which components should be open standards, and which should remain proprietary alpha?

In flexibility, those questions cut through the marketing. An API is not automatically a platform. A dashboard is not a strategy. AI language is not proof of differentiation. A virtual battery is a useful abstraction, but it still depends on physical availability, telemetry, settlement, grid connection and trust.

The anchor need: bankable flexibility under intermittent supply

The anchor need for this paper is not ‘battery software’. It is bankable flexibility under intermittent renewable supply. Bankable flexibility means flexibility that can be relied on, valued, financed, dispatched, settled, audited and repeated. That is a higher-order need than a single optimiser, VPP or battery project.

Different users express that need differently.

Map 1 - The flexibility value chain

Read more

This applies to business models and strategy as much as it does to technical ‘models’ and software. The ‘technical openness’ we want to develop must fit into a wider map of opportunity and strategy that can be developed in conversation.

A method that is becoming more and more established for conversations of this kind is Wardley mapping (or Value Chain mapping as it was once called). I haven’t so far seen Wardley Maps for the German energy domain. Now’s a good time to get the ball rolling.

I first saw Simon Wardley speak at an O’Reilly conference in London in 2016. (Here’s a link to the video of that event.) I won’t repeat his ideas here, and would recommend you watch one of his videos instead and make up your own mind on whether or not you think his approach is useful. (I think it is, and I do hope I’ll never have to see another SWOT matrix.)

Our first map will be high-level and focus on what is probably the overall goal. The analysis should be seen as a starting point, an illustration of the idea to be developed further.

We start with the map itself and then work through a Wardley-style analysis. I expect many comments and disagreements. Get in touch here on Substack, LinkedIn or otherwise.

1. Affordable, Reliable, Low-Carbon Electricity

Read more

The opportunities are not just technological. They are structural. Generative AI becomes more powerful when it is paired with open data and open implementations, when the underlying logic can be inspected, adapted, and extended.

To understa…

Read more

My agents chose to map the providers on two axes: ‘transparency/productisation’ on the horizontal axis and ‘BESS-modelling footprint’ on the vertical axis. Arbitrary? Yes, probably, but the simplified characterisations that emerge might act as useful foils for the fuller complexity. We always need to start somewhere.

The upper-right corner of our simplified map contains four very different approaches. Modo sits farthest right because it publishes detailed forecast and benchmark methodology, exposes an API, and runs benchmark governance as a regulated activity. Aurora is also strongly ‘productised’, but its public material still reveals more about the economics than the methods, assumptions, tools. AFRY combines a large storage footprint with some optimisation-engine detail. LCP Delta reaches the same corner by a different route: it describes a Europe-centred hybrid of forecasts, benchmarking, backcasting, data tools, and trading analytics. [Modo strategy] [Modo regions] [Aurora / TPG] [AFRY BESS] [LCP Germany storage] [Enact]

Baringa sits high on footprint but left of the most productised names because its public proposition seems primarily report- and judgement-led. It clearly has a serious model estate, including a pan-European PLEXOS-based model and a 31-market BESS Index, but it publishes less about the underlying storage engine than Modo, AFRY, or even Aurora. enervis sits lower and left not because it looks weak, but because it is intentionally narrower: its public storage proposition is still most convincing as a Germany-first specialist with a strong local power-market and policy heritage. [Baringa flex reports] [Baringa PLEXOS methodology] [Baringa BESS Index] [enervis history] [enerStorage Pro]

The academic literature helps explain why no single provider dominates every dimension. Reviews of storage economics and optimisation find no single preferred method across all use cases. Capacity-expansion and equilibrium approaches aim at preventing fantasy long-run rents; stochastic dispatch and trading models aim to capture realistic dynamics; myopic planning frameworks are probably more realistic about decision timing than perfect-foresight models. The commercial providers in this market are hybrids, but they still tilt toward one of those camps. [JRC review] [Hutchinson et al. 2026] [Butters et al.] [Fuso Nerini et al.]

Key findings at a glance

• Modo is positioned as a storage-first, benchmark-governed, API-accessible modelling platform.

• Aurora and AFRY offer a combination of power-market modelling depth and BESS-market scale, but they come from different intellectual traditions: Aurora from economics/software, AFRY from engineering/optimisation.

• Baringa has one of the broadest BESS market-coverage footprints, but its public trust mechanism looks more like ‘expert-led scenario judgement’ than ‘documented product architecture’.

• enervis is the most Germany-native of the core six and appears strongest where local market design, regulation, and co-location configuration matter most.

• LCP Delta sits in the middle: more productised and transparent than a classic advisory house, but still more consultancy-shaped than Modo.

Quadrant-style landscape

Axes: rightward means more publicly documented method and more productisation; upward means a broader and deeper BESS-modelling footprint. Positions are interpretive and based only on public ‘data’. Firms that most clearly productise and document their methods also tend to scale them geographically and commercially.

Read more

It seems to me that we’d want excellent causal analysis to get a sense for the interaction effects. The analysis should be rigorous, open-source, well-communicated, a basis for dialogue.

What we often have instead, it seems to me, are plausible stories built from timing coincidences, co-movements, and vivid failure scenarios. A starting point, perhaps, but no more.

Or have I missed something?

Read more

Calendar seasons are too ‘blunt’ for the German power market. They do not capture the move from a winter-compressed system into a spring market in which the middle of the day can soften sharply while the later system remains firmer.

…the recurrence of all sorts of days which not only nature, but adventitious circumstances, a purely conventional order, introduce into a season. Marcel Proust

A market-season vocabulary helps to identify the path more clearly than calendar seasons alone. In this report window, the useful sequence is Autumn Solar Fade, Winter Return, Winter Compression, and Spring Solar Ramp.

Seasonality in the German Energy Market

• Autumn Solar Fade: the high-solar period rolls away gradually rather than disappearing at a calendar boundary.

• Winter Return: solar falls back toward single digits and the system begins to feel winter-like again.

• Winter Compression: the deepest winter-like stretch, with weak solar and a system that still behaves as if winter is in charge.

• Spring Solar Ramp: solar is back strongly enough to soften the middle of the day, but the market has not yet settled into a summer plateau.

Summer Solar Plateau sits beyond this sample, but we will get there soon enough!

Between 2025-10-01 and 2026-04-08, as we’ll see below, the market moved from autumn fade into winter return, tightened into winter compression, and then opened into a spring solar ramp.

• Autumn Solar Fade (2025-10-01 to 2025-10-15): 15 days, average solar share 11.4%, average load 53.2 GW, dominant regime Surprise dislocation regime.

• Winter Return (2025-10-16 to 2025-12-31): 77 days, average solar share 5.4%, average load 57.1 GW, dominant regime Compressed baseline regime.

• Winter Compression (2026-01-01 to 2026-02-24): 55 days, average solar share 4.2%, average load 62.4 GW, dominant regime Compressed baseline regime.

• Spring Solar Ramp (2026-02-25 to 2026-04-08): 43 days, average solar share 20.3%, average load 56.9 GW, dominant regime Mixed transition regime.

…that fresh and piercing novelty of a recurring season, of a change in the routine of our hours… Marcel Proust

Winter gave us a Low-Low Baseline

Winter Return was the settling phase after the autumn fade: solar averaged 5.4%, load about 57.1 GW, and the dominant read was Compressed baseline regime. Winter Compression then deepened that structure. Solar slipped another 1.2 percentage points to 4.2%, load rose by roughly 5.3 GW to 62.5 GW, and the market stayed under the same compressed-baseline label.

And when I thought how I would see the return of that cold season which … had always seemed to me so melancholy, when I thought how evenings would come back like that snowy evening … what at such moments I still dreaded most…was the return of the intense cold, and I said to myself that what it would be hardest to live through was perhaps the winter. Marcel Proust

That winter baseline was also quieter in spread terms. Winter Return and Winter Compression were relatively contained SPARX regimes. Once the market moved into Spring Solar Ramp, both the average stress level and the upper tail moved materially higher. The spring market was not just brighter. In this sample, it left the market more exposed to intraday shape risk.

That is the baseline against which spring should be read. Once Spring Solar Ramp arrived on 2026-02-25, solar had climbed back far enough and load had eased far enough for the middle of the day to soften much more easily than it could in January.

Spring Repriced The Middle Of The Day

By late February, the market was no longer behaving like a winter system. Once it moved into Spring Solar Ramp, solar started reshaping the middle of the day more consistently. Midday pricing could soften much more easily, even while later-day tightness remained possible.

…a decanted springtime, which is reduced to its own essence and expresses the lengthening, the warming, the gradual unfolding of its days.. Marcel Proust

Three parts of that spring turn matter here. After 2026-02-25, solar moved into a higher operating range and stayed there much more often than in winter. Late March then wobbled without fully resetting the spring pattern: solar eased, load firmed, but the score layer stayed active instead of dropping back to the quieter early-February read. The Easter block belongs inside that same setting. Spring already looks like its own operating environment rather than a brief transition out of winter.

A useful late-March comparison window runs from 2026-03-24 to 2026-03-27, when the spring opening wobbled without fully resetting. The early-April holiday block in this report runs across Good Friday on 2026-04-03, Easter Sunday on 2026-04-05, and Easter Monday on 2026-04-06.

Early April Was A Shape Event

Easter Monday was a decisive break. It landed inside Spring Solar Ramp, with unusually weak demand, strong solar, and the daily classifier already in Surprise dislocation regime. DA average still sat near flat at EUR 3.04/MWh, but settled SPARX reached EUR 4274.16/MWh.

…from the muffling of sounds in the heat of the morning hours, the cool taste of cherries; the longing for Brittany or Venice from the noise of the wind and the return of Easter… Marcel Proust

Some of the setup was visible before Monday. By 2026-04-03, the classifier had already shifted into Surprise dislocation regime. By 2026-04-05, DA average was at EUR -16.34/MWh, and the 12-16 block had already turned sharply negative. A soft holiday midday was already in the day-ahead structure.

Day-ahead already showed a soft midday shape on 2026-04-06, but, in retrospect it does not capture the eventual scale of the break. 12-16 cleared at EUR -103.33/MWh, but 16-20 still printed EUR 30.50/MWh and 20-24 EUR 111.97/MWh. That left a 12-16 versus 20-24 day-ahead gap of EUR 215.30/MWh on the same curve. On the realised quarter-hour path, the intraday-minus-day-ahead spread still ranged from EUR -4233.90 to EUR 40.26/MWh across the day. That is why Monday produced the largest SPARX break in the sample.

Read more

Energy systems concentrate public-law duties, private investment, market coordination, cross-border interdependence, and decarbonisation pressures in one operational field. It is a system of constant negotiation across public and private systems. What could be more fascinating as an object for study and discussion?

The German setting is especially interesting, I believe. The Energiewende has led to a slow-motion unfolding of complexity as high-level commitments meet questions of supply security, affordability, and the practical need to translate those commitments into grid expansion, planning, redispatch, connection, and balancing rules.

To get closer and explore the dynamics, let us now suppose that it is a “day in the life in the future.” AI agents are being widely used to solve problems. How do we imagine this would be working itself through in energy systems and settings?

ChinaTalk

Claude Just Opened the Strait

In what analysts are calling “the most productive jailbreak in diplomatic history,” Anthropic’s Claude model reopened the Strait of Hormuz early Sunday morning. This shocking development came hours after President Trump threatened to obliterate Iran's power plants if the strait wasn't reopened within 48 hours, singlehandedly preventing global recession…

Read more

2 months ago · 943 likes · 193 comments · Jordan Schneider

How do public-law duties, critical-infrastructure rules, regulator determinations, operator obligations, market rules, contracts, and judicial review become harnesses for automated systems to determine who may decide what, on which grounds, and subject to what form of challenge?

An agent designed to operate under constitutionally relevant constraints would need to recognise the legal basis of its task, the non-negotiable limits on its role, the priority rules between relevant layers, the points at which uncertainty must be escalated, and the need to preserve reasons for later review.

1. Operational authority in a critical system. Redispatch brings public system aims and private asset operation into the same legally structured decision space. It can both constrain preferred private operation and recruit flexible private assets as a service to the grid. In March 2026, TransnetBW and Octopus Energy reported redispatch processes using more than 700 household electric vehicles. The question for agentic support, if introduced into such a setting, is what may be recommended, what may be triggered automatically, what requires human confirmation, and which layer governs when security of supply, cost efficiency, and decarbonisation pull in different directions.

2. Priority-setting in connection and buildout. AgNes and the 2026 Reifegradverfahren for large connection requests show how scarce capacity is allocated through readiness criteria and system-benefit considerations, with direct consequences for buildout timing and contract structure. The question is who defines the optimisation target, and how national efficiency, local affordability, and regional fairness are translated into operative priorities.

3. Cross-border collision. EU balancing and forward-capacity-allocation rules may meet German statutory duties, TSO constraints, vendor model policies, and firm-specific risk controls in the same workflow. Interconnectors and allocation rules can widen the space for trade and balancing across borders, but public TSO materials also make clear that cross-border capacity remains physically constrained and jointly allocated through market mechanisms. The question is how such collisions are mediated: by hard override rule, negotiated escalation, interface design, or technical default.

Taken together, these three questions prepare the transmission-chain analysis that follows. Full article and source links at the end of the post.

Constitutional Transmission in Germany

The German energy case shows how constitutional commitments reach a dense infrastructural setting of regulators, network operators, private contracts, and software. The claim is not that the Basic Law directly governs every operational decision. It is that constitutional commitments are translated, through increasingly specific legal and institutional forms, into operative permissions, constraints, procedures, and review structures. That sequence is the transmission chain that becomes increasingly relevant to our imagined agents.

1. Constitutional principles. At the top sit basic constitutional commitments: democratic legitimation, the rule of law, administrative binding to law, and, increasingly, environmental orientation. These do not tell a TSO or DSO how to run redispatch on a Tuesday afternoon. They do, however, define the conditions under which public power may be delegated, constrained, justified, and reviewed.

2. Statutory concretisation. The Energy Industry Act carries those higher commitments into sectoral public purposes and competence structures. It gives the energy order a legal telos, allocates responsibilities to transmission and distribution operators, and creates the legal basis on which more detailed regulatory ordering can occur.

Read more