At this year’s Smarter E Europe trade show in Munich, Envision Energy’s Solutions Director for Europe and the UK, Michael Koller, framed the company’s pitch around a single integrated machine: solar generation, battery storage, grid-forming inverter capability and AI-based optimisation engineered as one product rather than four separately procured components. The timing matters because grid-forming behaviour is moving out of pilot projects and into actual connection requirements on European grids, and the solar-plus-storage plants now being contracted are the first wave that will have to deliver it. A developer that still treats solar, storage, inverter controls and optimisation software as independent vendor decisions is not just adding integration cost – it is locking in a plant that may not pass the stability tests the next grid code version asks for.
The market reality of hybrid plants built as one dispatchable system
The background Koller is speaking into is a European market where solar-plus-storage co-location has stopped being an optional pairing. Network constraints, negative price hours and tightening permitting rules have pushed batteries from an add-on feature to a structural part of new solar capacity. In most of these hybrid plants, the technical thinking still comes down to a standard arrangement: PV inverters follow the grid, the battery inverter follows a local converter, and a plant controller blends the two. That arrangement works while a large synchronous grid holds frequency and voltage fixed. It goes talk upstream closer to grid-forming, where the battery itself has to set the reference for the local network.
Grid-forming is a specific technical change. Conventional grid-following inverters need an external voltage/frequency reference – they lock onto the grid like a phase-lock loop and obey it. Grid-forming converters generate their own internal reference and can behave as something that resembles a synchronous machine: they provide short-circuit-current-like inertial response, they can keep a local island energised after a disturbance, and they can restore frequency in seconds rather than turbines spooling up. The price of that capability is that control loops can interact dangerously if multiple grid-forming units run incompatible strategies. That is precisely why Koller’s “one system” raises such a practical safety case in solar plants with many PV inverters plus a large battery connected at one point of couple. If the battery grid-forming unit has to respond to solar inverters bridging through the same collection grid, the enforceability and that sequence is where is the set of voltage-voltage dimension.
There is also AI. In Envision’s picture, AI is not a dashboard that predicts one day; it is the layer that joins the physics of the plant with the commercial dimension of dispatch. It predicts solar ramps, assesses storage degradation, tracks a dozen market products deliverable concurrently, and, above a threshold, takes solar/storage dispatch decisions in the seconds that grid services require. The reason that needs to be developed along the same model as the inverter and the battery is that a gripper-controlled interconnect will pass signals at grid speed while an AI stack is working at schedule speed; any communications mismatch – separate controllers, different vendor protocols, mismatched time bases – breaks the coordination case.
What is not small is the integration burden this creates. Purchasing four components from four vendors means paying for compatibility engineering, writing gap agreements between warranties, and sitting at test with a grid harmonic or modal that is the permanent home of many projects. The integrated argument – one The responsible party for the plant procured, all, is about the same as a relational issue.
Converging currents: grid-forming mandates, co-location economics and the AI trading branch
The interesting part is that three separate trends now collide. The first is the official move toward grid-forming. Australia’s System Operator AEMO has already forced grid-forming behaviour into prescribed connection requirements for large batteries – the market is now the most advanced utility-scale grid-forming ecosystem that exists anywhere. Australia’s South Australian neighbouring grid is a research lab for it; most European operators started from stability-services contracts rather than, a slowly-coming threshold. In Great Britain, the former National Grid ESO went a series of stability “Pathfinder” contracts that allow batteries to sell inertia and the equivalent. On top of network operators have begun to question. It is not far from here: with explicit NFC technical requirement for grid-forming in the connection code – the trend is plainly toward mandating it at large BESS new grid.
Second streak is co-location and revenue. Solar plus storage already widely makes sense only when generation and storage dispatch to a multiplicity of as-day-ahead, frequent, capacity, FCR and wider total times. But the margins for these uses are compressing as more capacity arrives, so the stack of optimal operation is longer. That naturally brings in the AI trading layer: Some AI Tovisoftware can re-forecast the solar plant with satellite imagery, forecast, storage performance accounting accelerating, and make decisions per 15-minute market products. The winner in this market will be the one that can bundle that algorithmically-driven dispatch inside the actual plant controller, so the dispatch is not delayed by a central operator and it can be sent directly to the inverter sets, including the grid-forming unit.
By comparison, – here I off. That means grid-forming is a premium product studied in a 2- to 3-year lifecycle. Third, the storage industry plans into one, Chinese and Korean suppliers enter into a single supplier model – effectively storage system ownership of the plant controller against integrators or best-integration. If this trend holds, the next wave of hybrid auctions are likely to become fewer but larger vertical platforms; European and United more choices go down; commissioning time may be shortened, but exit costs, if you change supplier, will rise proportionally much more drastically.
Where does that leave the conventional deployment of pure hardware? If grid-forming clarity matters at the level of millions of MW of generation, then an imposed or the retrofitted function for grid stability tends to be measured in watch turtle layouts – for an inverter, a 100 MW, maybe 250 to 450, layered onto by a synchronous generator – the mechanical control integration – suggests a project team sees a few to values – but a purely third-party integration is not a thin task; it’s a foster-year. If every further grid code formerly as an equivalent of 15 pert – that would pull total capital cost – compression…
All the while, an external contradiction: the AI piece needs open data to be effective, and integrated vendor stacks rarely use fully open interfaces. Envision’s unified full stack requires client data – plant logs, temperatures, SOC and forecast – inside their optimised engine. For a plant in a PPA, that data belongs to asset manager or the asset owner; the data flows are tuned to the economics. That suggests an answer is likely a fourth integration monetised: not an open “the grid”, but an “AI inside the plant” – and whether the controller also exports the same AI output outside the door of the vendor model.
Who this affects: decisions that change at procurement and operation level
- Utility planners and system operators – planning can no longer assume that hybrid PPP the utility stack arrives with “grid form” as a technical guarantee. Planning the connection you under a full-stack hybrid that has already passed its own PV+storage stability simulation, giving faster, cheaper study, at the price of no longer being able to independently define the interior plant architecture.
- Solar and storage developers – you now have to agree on the choice as filter: a short list of integrated platforms (solar + battery + a controller that attacks grid-forming via advanced inverters) versus opening best-of-suite components. If you keep best-of-suite, add mandatory performance testing to the RfQ of a grid-forming BESS, and execute with a single integrator responsible on P77. The cost of a late retrofit – swapping an inherited or where integrated controller – is permanent, that stress of a bad half-year.
- Grid code and policy analysts – be careful not to write grid-forming. Codes demanding “grid forming, period” have proven to be the most flexible way to deliver successfully by manufacturers. The healthy requirement are in bandwidth of dl/dt, voltage support response, and dynamic data sharing. If you instead ends up describing just one particular company plan, you’ve subsidised their / de facto physical use: you decided by tech standard; otherwise there’s a risk that a “AI-slight” controller-deep creep in market rules.
- Investors and asset managers – during the buying decision stage, assign value to the open, extensible architecture of the platform. The resale value of a hybrid asset is a cost trade-off: an integrated stack is marginally lower construction risk, but to float it from capex/value; worth strict – legacy differences in circuit side about this. On a 10-year operating case, the difference between an integrated AI layer and a third-party – want forecasts and different can’t duplicate – is a pivot influencing decade-long returns.
What to watch next: the signals that verify integration
- European reference project name: track when energy Enso reports its first European or UK full-stack hybrid hybrid with a clear term: “solar + storage with grid-forming and GV AI.” At first, in a distribution; only the first few are published names; if delivery slips due to wider phase relation, see code with a higher risk.
- Grid code progress in Europe – take on ENES-E (or national H1 grid) that explicitly defines the phrase “grid-forming” in RFC / DRU term. Watch for a first technical draft of defining the usability test – not a participation criteria. Number time: it will affect the very stack of existing hybrid plants.
- Product announcement on “time to milstone”: watch how well concedes one global updates: peacetime – architecture / digital twin; – particular for the polynomial. If a vendor starts publishing AI-detailed results (PV models tested vs P executed), that’s the first visible sign that a “integrated system” is actually in operation, with the grid numbers.
- Commercial balance: – list of projects 1-2 years on how many French hybrid plants beyond Germany, some prototypes (~20-30 MW), and separate, vertical. If you’d force together usually head > 1 × as simple..; if the rational tests make the integrated frame cheaper goods, the purchasing cascade is beginning.
Bottom line
The next contract for an integrated solar-plus-storage plant is not really about solar panels, batteries, inverters, or “machine-learning” – it is about which supplier you design the entity grid control model and its dispatch decision process – from the PV ramp to the market bid. Envision’s argument that it should be a single owner in that process is not surprising given the face, but it is a signal to the rest of the industry: decisions that were promised in order to be changed into four independent buyers are about to become a single control architecture choice. The developers making the decision the fastest with a consequence of weight will be those who demand – regardless of the vendor – that the grid-loop, the battery loop, and the optimisation loop are one coherent one.
Read the full report at Energy Storage News
Note: facts and figures attributed above to Energy Storage News reflect that outlet's original reporting. Broader context, cross-sector connections, and forward-looking scenarios reflect independent analysis by our editorial team.
About this article: Drafted by Energy Ai with AI-assisted research and writing based on public reporting, then reviewed under our editorial process before publication.
Leave a Reply