Turning Buildings Into Grid Service Providers: Evidence from Two Swiss Pilots

The growing case for flexibility at the grid edge

The distribution grid was designed around a fairly predictable pattern: centralized generation feeding consumption that varied within known, predictable bounds. Reality is already quite different today, and generation and consumption will continue to change in the coming years. Local solar generation, heat pumps, batteries and electric vehicle (EV) charging are shifting load and injection profiles toward peaks that are larger and harder to forecast than the ones the grid was built for, because of renewable volatility. Switzerland's own regulatory response reflects the scale of the shift: a new flexibility regulation, in force from 2026, gives suppliers and consumers new options to make greater use of flexibility for the benefit of the grid [1].

Across Europe, the increasing grid congestion is no longer marginal either: a May 2025 study by Beyond Fossil Fuels, E3G, Ember and IEEFA found that grid capacity constraints led to an estimated EUR 7.2 billion in curtailed renewable generation across just seven European countries in 2024 alone, with roughly 1,700 GW of renewable projects still waiting in connection queues [2].

Reinforcing grid infrastructure is one response, but it is capital-intensive and slow relative to the pace at which distributed assets are being installed. The complementary lever is demand-side flexibility: enabling buildings and EV chargers to shift when they draw or inject power, so that existing grid capacity can serve more demand without local bottlenecks. Demand-side flexibility is increasingly recognized as an alternative to reduce system costs and ease congestion without new grid infrastructure investment [3,4].

Coordinating three roles: grid operator, building operator, orchestrator

Making demand-side flexibility available in practice requires connecting three parties:

• Grid operators know where and when local constraints occur.
• Building and fleet operators control the distributed energy resources, i.e., batteries, EV chargers, heat pumps and solar panels, that could respond and are ultimately responsible for the comfort and service level their own users expect.
• An orchestration layer sits between them: forecasting demand, optimizing local dispatch, and translating a grid operator's request into a control action, and back again into a verifiable signal the grid operator can rely on.

The two pilots summarized below, one conducted by kuafu together with Energie 360° and Smart Energy Link, the other by kuafu with Energie 360°, ewz and SmartGridReady, were designed to generate the following evidence: first, that a building operator's energy resources can be optimized to respond to grid tariffs without degrading the service its users receive; and second, that the resulting demand-side flexibility can be reported to, and constrained by, a grid operator in real time to manage congestion.

Evidence from the field

In the first pilot, a collaboration between kuafu, Energie 360° and Smart Energy Link, two sites were used. At the Arrivo multi-family development in Kloten, a 40 kW / 80 kWh battery combined with 120 kWp of solar was operated with kuafu's predictive control engine, targeting self-sufficiency and peak shaving, over a test period from June to September 2025. Smart Energy Link provided the device infrastructure for kuafu to communicate with the devices in the building, supplied real-time data on the site's energy balance, consumption and production, and was in charge of safe operation in case of communication loss.

Figure 1 illustrates the resulting operation over a single representative day: kuafu's predictive engine forecasts the site's load profile (red curve, top figure), including its evening peak, and pre-charges the battery (orange line) earlier in the day, as soon as solar generation is available, to maximize self-sufficiency. Conventional peak-shaving controllers typically rely on a fixed threshold that has to be pre-configured; kuafu's engine instead forecasts the peak in advance and times the battery's discharge automatically, reducing grid electricity imports. Measured against the standard benchmark controller, this improved cost reduction by 72% in the test period.

At the second site, Energie 360°'s headquarters in Zurich, the corporate commercial fleet with 55 EV chargers operates under a dual e-mobility tariff (Rp 25.05/kWh peak, Rp 5.65/kWh off-peak). The same predictive engine was applied to charging schedules over a test period from August to September 2025, with the objective of grid tariff responsiveness. Figure 2 below illustrates the resulting operation over a single representative week: rather than throttling demand across the board, the engine shifts individual charging sessions out of high-tariff hours while still meeting every user's charging demand in full. Measured against the same rule-based benchmark (blue bars), kuafu's engine-controlled schedule (red bars) reduced charging costs by 11% over the test week shown. At the time of testing, this site operated without on-site solar and under a fixed two-tier tariff. With on-site PV generation and a 15-minute dynamic tariff, kuafu's predictive engine would result in significantly higher savings. The important outcome is that the user charging experience was not impacted.

Both results speak to the building operator's side of the equation: predictive control can reduce cost without reducing the service level users experience.

Closing the loop with a grid operator: the Flex2gether pilot

The second pilot was a collaboration between kuafu as the orchestration platform, ewz as the grid operator, and Energie 360° as the asset owner, specifically the e-mobility assets at its headquarters. The objective was to define the first SmartGridReady functional profile for an orchestrator platform like kuafu, which did not previously exist, and to test it in a real setting, enabling the grid operator to procure local flexibility. A flexibility signal is only useful to a grid operator if it can be trusted to hold under real operating conditions, and if it is exchanged in a format the grid operator's own systems can process.

The test setup was as follows: ewz sent a site-level power limit the day before, and kuafu controlled the site's EV charger load to stay below this limit while protecting users' charging demand in full. Because no orchestrator-to-grid-operator interface of this kind had been tested previously, kuafu worked with SmartGridReady to define a reusable functional profile specifying the data points exchanged in each direction, enabling the procurement and provision of demand-side flexibility.

The test was run over five business days. Figure 3 shows the result under a fixed 80 kW ceiling communicated by ewz the day before (red line): without control, the site's own charging pattern would have exceeded the limit for an estimated 420 minutes and 72.7 kWh; under kuafu's active control, exceedance fell to 45 minutes and 3.3 kWh, a reduction of roughly 90% in time and 95% in energy. The red-shaded area shows the load that kuafu deferred relative to the uncontrolled baseline, and the green-shaded area shows that same energy being released once capacity became available again; the site still delivered 2,044 kWh to users across 112 charging sessions.

What the pilots suggest, and what comes next

Together, the two pilots close the loop described at the outset. The Arrivo Kloten and Energie 360° headquarters pilots show that a building operator's own energy resources can be optimised for energy cost savings without degrading the service its users receive. The Flex2gether pilot shows that consumption can be shifted to reduce the grid load, in response to a power limit that ewz communicated a day ahead and that kuafu's platform enforced in real time, using a standardised interface rather than a bespoke integration.

Extending this from one site to a portfolio of sites, and from one grid operator to several, is the relevant next step for kuafu, alongside the commercial and regulatory question of how a building operator or asset owner that provides this service should be compensated for it. Switzerland's own grid strategy already gives this a name: under the NOVA principle (grid optimization before reinforcement before expansion), enshrined in the multi-year network planning obligations of Art. 8 of the Electricity Supply Act (StromVG), demand-side flexibility is not an add-on to grid planning but the first lever grid operators are expected to draw on before physical grid reinforcement or expansion is considered [5].

The two pilots are evidence of what optimization and orchestration can concretely look like at the level of a single building and a single grid operator's constraint, and of the benefit this brings for grid decarbonization. kuafu's own ambition follows from that reading: to make demand-side flexibility from decentralized energy resources available to Swiss grid operators as a standard planning tool, not a pilot curiosity, and to make the buildings that provide it fair partners in that exchange rather than passive sources of congestion.