XL BESS Projects in Germany – Article 2: BESS Contracts as a Neural Network & Why Individually Good Contracts Can Still Sink a Project

After more than 8 GWh of executed BESS projects across Germany and Saudi Arabia, the part of project delivery I find most analytically interesting is not the drafting of individual contracts — that is the lawyer's craft — but the architecture: how the contracts in a portfolio fit together, and what happens when one pulls in a direction the others were not designed to follow.

The most expensive disputes I have seen came not from badly drafted contracts, but from individually excellent ones that nobody asked to coordinate. This article uses three scenarios from recent BESS operations to propose a different framing — the contract portfolio not as a stack, but as a network — and draws practical conclusions for what that means at signing and in operation.

Scenario 1 The availability squeeze (most common scenario)

A 200 MW / 400 MWh project in Germany delivers less availability than promised. The LTSA and the SLA do not guarantee a restoration time, which is the case for the majority of agreements I have seen in the market, and the spare parts needed for the failed components are on long lead. Availability slips below the threshold. The owner now sits in the middle of six counterparties at once.

The OEM warranty covers the component but not the consequential loss. The LTSA does not commit to a restoration deadline. The SLA contains soft KPIs that do not translate cleanly into recovery. The Optimizer (Vermarkter) begins applying imbalance and non-delivery penalties against the direct marketing contract. The toller does the same against the capacity envelope. The lenders observe the availability dip approaching the covenant trigger and begin asking questions about the DSCR. If the asset is committed to critical grid services such as black start, the TSO becomes the seventh counterparty, with its own prequalification status and contractual penalties at stake.

The root cause of this trap sits two phases upstream, in engineering. The spares strategy, the restoration time commitments, and the redundancy assumptions were either not specified tightly enough or not flowed back into the contractual envelope. Quality is the first line of defence for exactly this reason, but once the gap surfaces in operation, the penalties flow outward to the commercial counterparties and the conversations with lenders become uncomfortable.

Each of the six or seven counterparties is performing as written. The owner is the only party absorbing the residual.

Scenario 2: The degradation dispute

Eighteen months after PAC, a different 200 MW asset begins showing a degradation signature the LTSA contractor classifies as accelerated wear. The owner files a warranty claim. The LTSA counterparty responds that the dispatch profile, driven by the Vermarkter under the direct marketing agreement, falls outside the duty cycle assumptions baked into the warranty. The Vermarkter, in turn, points to the tolling-style optimization rights granted in its contract and the EMS setpoint authority embedded in the OEM scope. The Operational All Risk (OAR) insurer reviews the file and invokes the utilization clause, reducing settlement. The financing agreement contains a covenant linked to availability, which is now under question.

Four contracts say it is not their problem. Each is internally coherent. The owner absorbs the loss.

Scenario 3: The shrinking grid envelope (FCA is comming)

A 100 MW / 200 MWh BESS in Germany is energized under a Flexible Connection Agreement, signed two years earlier as the only way to bypass a grid connection queue that now exceeds 720 GW of pending BESS requests against a national peak load of around 80 GW. The agreement caps import-export power under congestion, imposes a ramp rate restriction, and limits the share of capacity prequalifiable for FCR and aFRR.

Eighteen months into operation, the DSO invokes the curtailment provisions more aggressively than the business case assumed. Annual revenue drops by around 20 percent. The Vermarkter, who priced the offer against an unrestricted dispatch envelope, asks to renegotiate. The toller does the same. The LTSA performance regime, written around full nameplate availability, no longer maps onto the asset's economically usable capacity. The financing model drops several percentage points of IRR and approaches the covenant threshold.

Every counterparty is exercising rights it negotiated. Every contract is internally coherent. None was wired to handle a downward shift in the regulatory loss function.

I had not seen this failure mode in earlier projects. Capacity-reducing interventions by the DSO post-energization were almost unheard of under the old regime, where capacity was firm and curtailment was compensated under the EEG. That world is ending. With FCAs becoming a standard mechanism for grid access and the German government's draft reform pointing the same direction, the regulatory frame around BESS projects is moving inward — and most contract portfolios were not designed for this.

What all three scenarios have in common

Six contracts, four contracts, or an entire portfolio rewired by a single FCA invocation. Each agreement in each scenario was negotiated by competent counsel. Each, read on its own, is internally coherent. And yet in every case the project ends up in a place where no single counterparty is fully responsible, no single contract fully resolves the issue, and the owner is absorbing the loss while disputes proceed in parallel.

This is not a contract drafting problem. It is a contract architecture problem. And it is best understood by changing how we think about what a BESS contract portfolio actually is.

The shift in framing

Most discussions of BESS contracts treat them as a stack: development agreements, then EPC, then LTSA, then O&M, then the commercial agreements layered on top. The mental image is sequential and hierarchical. Each contract is reviewed, negotiated, and signed in isolation, often by a different team, at a different phase, with a different counterparty on the other side of the table.

But that is not how the project behaves once it is energized. Once the asset is operating, the contracts are not stacked. They are connected. Signals flow between them constantly: telemetry, performance data, dispatch instructions, warranty claims, availability reports, insurance notifications, regulatory filings, financial covenants. Each contract is a node. The interfaces between them are weights. And the intelligence of the project, meaning its capacity to actually deliver what the business case promised, does not live inside any single contract. It emerges from how the network is wired.

A network with the wrong weights produces wrong outputs even when every neuron is individually excellent. The same is true of a BESS project. A perfectly drafted EPC and a perfectly drafted LTSA, connected through an interface that does not define availability the same way, will produce a dispute. Not because either contract was bad. Because the weight between them was wrong.

The entire network sits inside a regulatory framework — grid codes, connection rules, cybersecurity and resilience regulations, environmental and permitting law. Regulation is not another node. It is the external function the whole network is being optimized against. When regulation changes, the network needs retraining, which is the polite name for renegotiation.

How the network actually behaves

Think of a real BESS project network. The EPC node is connected to the OEM supply node by weights that carry technical specifications, factory acceptance criteria, and delivery schedules. The OEM supply node is connected to the LTSA node by weights that carry warranty terms, spares logic, and performance baselines. The LTSA node is connected to the Vermarkter node — even though they have no direct contract with each other — through the asset's operating profile and the duty cycle the EMS actually runs. The Vermarkter node is connected to the toller node through the dispatch authority and the optimization envelope. The financing node pulls weight from every other node in the network through the covenants it imposes, the direct agreements it requires, and the step-in rights it reserves. The insurance nodes — CAR, EAR, DSU, OAR, Business Interruption Insurance (BI), cyber — sit alongside, with weights into almost every other node depending on which phase the project is in.

What makes this a network rather than a list is that a change at one node propagates outward. A delayed shipment under the OEM contract changes the EPC schedule, which moves the PAC date, which shifts when OAR attaches and when DSU stops paying, which affects when the LTSA performance regime starts, which changes how much of the cell warranty has already burned before commercial cycling begins. A single delay touches six contracts. None of those contracts contains a clause that describes the propagation. The weights between them carry the propagation silently.

And what makes the network intelligent or unintelligent is the quality of those weights. A well-wired network handles propagation without producing disputes. A badly-wired network produces a dispute every time a signal crosses a misdefined interface.

The interfaces where misalignment is most expensive are the ones that connect contracts that were negotiated by different teams, at different phases, with no common drafting reference. Those are the connections the rest of this article diagnoses.

Interface failure one: dispatch authority

The Vermarkter has commercial authority to optimize within the technical envelope. The EMS executes setpoints. The LTSA prices its guarantees against an assumed duty cycle. The prequalification was issued against specific measurement points. The balancing group settles against the meter operator's reads. Five contracts. No shared definition of dispatch authority, duty cycle, telemetry reference, and settlement boundary.

The LTSA counterparty logs cycling that exceeds the warranty envelope and starts building an abnormal-use defense for year three. The TSO observes response drift from the prequalification baseline, putting status at risk. The balancing group accumulates imbalance penalties because the EMS and the meter operator are on different time bases. The Vermarkter keeps optimizing for intraday spread because nothing in its contract tells it not to.

Each contract performs exactly as written. The network fails because no contract shares a common operating definition of what the asset is being asked to do. A single performance taxonomy and a single dispatch authority document, referenced by every contract that touches operation, would change the economics of the project more than any individual clause negotiation.

Interface failure two: the warranty and insurance donut hole

This one I have lived through personally.

Cells leave the factory in China under a marine cargo policy. They arrive at the port and transit to site, often passing through a period of storage where the cargo policy has expired and the CAR policy has not yet attached because erection has not started. The cells are installed under EAR, which transfers to OAR at PAC. The chemistry inside the cell ages calendrically from manufacture regardless of cycling, and the OEM warranty clock began at delivery or two to three months after delivery. The LTSA performance regime began at acceptance. The DSU policy pays for delayed revenue only if there is an admitted physical damage claim under one of the underlying policies.

Now picture a container damaged in a storage yard at site two months before erection begins. If the contracts were not deliberately wired to close the transit-to-storage-to-erection chain, the loss may fall outside every policy in the network. The DSU recovery then collapses too, because DSU is parasitic on the underlying physical damage cover. The financing model assumed DSU would protect revenue during construction delays. The project is now exposed.

A milder version of the same failure repeats at PAC. The cell has been aging since manufacture and the OEM warranty has been running since delivery, often many months before PAC. The LTSA only begins warranty intermediation at acceptance. The OAR attaches at PAC. If commissioning slipped, the cells may be a measurable fraction of the way through their warranty life before they ever cycle commercially. Nobody flagged this at signing because each contract was reviewed in isolation against its own start date.

The interface that needs to be wired here is not a clause in any single contract. It is a coherent insurance and warranty timeline, mapped against the actual construction schedule, with explicit responsibility for the transitions between policies and warranties allocated to a named party. This is integration engineering applied to contractual risk transfer, and it is rarely done well.

Interface failure three: the duty-cycle insurance trap

The most under-discussed failure mode in the industry and, in my experience, the most destructive of equity returns in year three to five.

A tolling agreement, a direct marketing agreement, or an aggressive ancillary services participation transfers operational authority to a commercial counterparty. That counterparty's job is to maximize value within its contractual envelope. The LTSA priced its warranty against an assumed duty cycle. The OAR underwrote the asset against "normal operation" with no quantitative bounds. The BI cover follows the OAR.

When the commercial counterparty cycles harder than assumed, the LTSA builds an abnormal-use defense and the OAR insurer invokes a utilization clause to reduce settlement or void coverage. The owner ends up uninsured on a degradation event they did not cause and could not control, because operational authority sits with a counterparty that the LTSA and the insurers were never asked to coordinate with. Equity absorbs the loss. The financing model never modeled this scenario.

This is the cleanest example of how the network metaphor changes the diagnosis. Looking at any single contract — the tolling agreement, the LTSA, the OAR — none of them is defective. The defect lives in the unwired interface between them, and that interface only becomes visible once you draw the contracts as a network and look at what is missing between the nodes.

The cybersecurity perimeter exhibits the same pattern and is the subject of an earlier article in this series — a clean illustration of how an obligation that sits squarely on the asset owner gets distributed across four or five contracts that were never asked to coordinate with each other.

Backpropagation, or what claim management actually is

When something fails in operation, the cost has to travel backward through the network to find the misaligned weight. This is what claim management actually is — backpropagation through a contract network. Most BESS networks were not designed to carry the backward signal cleanly. The forward path was negotiated; the reverse path was not.

That asymmetry is why disputes in this industry are so expensive and so slow. The backward signal hits unspecified interfaces and weakens at each one. The owner absorbs the residual.

A network designed with backpropagation in mind has explicit loss-allocation logic at each major interface, shared definitions that survive a dispute, step-in rights that let the financier intervene at any node, and insurance policies with coordinated definitions of incident, cause, and consequence. None of this is more expensive than the alternative. It is dramatically cheaper, but only if it is built in at signing rather than reconstructed at year three under dispute.

Practical advice from someone who lives with these problems

Contract management has fascinated me since my PhD, and it remains one of my favourite parts of any project. From experience, what works:

One , keep the contractual framework in your head as a neural network whose only purpose is the smooth execution and operation of the project. No contract and no phase is an island, and the moment you start treating any single contract as self-contained, the network begins to fail silently.

Two , build a Division of Responsibilities (DoR) matrix for the operation phase that reflects your revenue stack and coordinates every party in the network, including those who have no direct contract with each other (the toller and the LTSA contractor, the warranty provider and the Vermarkter). Then run operational scenarios against that matrix. Force the scenarios to expose the communication paths between parties who have no contractual relationship but who depend on each other for the asset to function.

Three , those scenarios will surface the dependencies and the gaps that the asset owner ends up filling. Some of those gaps can be closed in engineering and project management. Others can only be closed financially, through contingencies and reserve accounts. Either way, you now know where they are, which is the precondition for managing them.

Four , accordingly build a similar DoR for the design and execution phase. I use it as the base for any contractual negotiation, and I attach the relevant section to the RfP or RfQ of each supplier. Suppliers will rarely give you everything you ask for, but the matrix lets you track every change and preserve the harmony between elements that would otherwise drift apart in separate negotiations.

Five , build an Operating Profile appendix for the station that travels with both the supply agreement and the operation contracts. It is a single document summarizing the engineering envelope, the warranty boundaries, the LTSA regime, the diagnostic data flows, the Vermarkter rules, the Technische Betriebsführer scope, and the dispatch authority logic. It is one of the cleanest tools I know for keeping operation inside warranty limits without forcing the operations team to read every contract every week.

Last but not least, any investor planning an XL or several XL-scale BESS stations needs a dedicated asset management department, staffed with engineers educated and trained to manage these assets and coordinate between the technical, commercial, and finance functions during operation. This is a job no one else will do for you, and no contract structure will substitute for having competent people inside the owner's organization who see the whole network.

Summary

A BESS project is not a stack of contracts. It is a network. Individually excellent contracts can still produce a failed project if the network is wired badly, and the wiring decisions are made at the interfaces, not inside the contracts themselves.

The next time you review a BESS contract portfolio, try this. Print the contracts as a list. Then draw them as a network. Mark the heavy interfaces. Ask which interfaces carry the most weight, where the connections are missing, whether the backward path has been specified as carefully as the forward path, and whether the network would still function if the regulatory boundary moved inward. The answers will not be in any single contract. They will be in the spaces between them.

Originally published on LinkedIn Pulse. Open the original →