Digitalization and Smart Grid

The digital and software layer that turns a passive distribution grid into an actively managed, flexibility-enabling system. “Smart grid” (smarta elnät) is the umbrella term for the metering, control, monitoring and data-exchange capabilities that let DSOs and TSOs see grid state in near-real time and act on it — the precondition for Congestion Management through Flexibility rather than copper. This page is an orientation hub; specific mechanisms have their own pages.

What “smart grid” covers

Swedish monitoring — Ei’s biennial smart-grid report

Under Electricity Directive 2019/944 Art. 59.1(l), Ei now monitors smart-grid development biennially. The first edition, Ei R2026:02 (December 2025), uses 13 indicators across three categories — Förutsättning (preconditions, e.g. installed local production, connected storage capacity), Användning (use), and Prestation (performance). Findings are tentative (only two years of data). See also the underlying Source - Ei SGI Data 2023-2024.

EU-level framework — CEER’s smart-grid performance indicators (January 2026)

CEER’s follow-up paper to the June 2024 ACER-CEER guidance (Source - CEER Electricity Smart Grid Performance Indicators (2026), published 8 January 2026) proposes 14 output indicators across 6 distribution-relevant dimensions for NRAs to monitor DSO smart-grid performance, implementing Art. 59(1) of Directive 2019/944:

Not mandatory or fully harmonized — CEER’s position is that only a few core indicators should be common across Member States, with a wider optional basket left to national discretion. Sweden’s existing Ei R2026:02 (13 indicators, three categories) predates this EU framework by a few weeks; whether/how the two will be reconciled is an open question (see data gaps).

EU legislative track — smart grid indicators and data exchange in COM(2026) 600

The Commission’s July 2026 proposal to amend the Electricity Regulation (COM(2026) 600) would turn smart-grid monitoring into an EU process. Art. 18a(2)–(3): ACER, with the Commission, ENTSO-E and the EU DSO Entity, would issue a recommendation on smart electricity grid indicators within 12 months of entry into force, TSOs and DSOs would supply the data to regulators and to ACER, and ACER would publish a progress report at least every three years; the Commission could adopt the indicators by implementing act (Art. 61(5a)). Art. 18(2)(a) would also include the indicators among the performance indicators regulators set. Art. 18a(4)–(5): TSOs and DSOs must exchange grid data in a harmonised manner under the Data Act, and ENTSO-E and the EU DSO Entity must set up a voluntary secure grid data exchange framework for research and innovation within 12 months, with implementing acts to follow (Art. 61(5b)).

The EU DSO Entity supports the data framework but asks for more than 12 months and coordination with parallel initiatives; it opposes folding indicators into efficiency benchmarking, the Commission empowerment on indicators, and direct DSO reporting to ACER, arguing that the national monitoring already done under Art. 59(1)(l) of Directive 2019/944 (Sweden: Ei R2026:02 above) is the right level (DSO Entity’s reaction). How the Commission indicators would relate to CEER’s 14 indicators or Ei’s 13 is not addressed in either source.

R&D direction

Svk’s research programme (Uppdrag 3.5, 2026) names digitalisation as one of four R&D areas, with the report’s concrete examples including AI/ML forecasting as control-room decision support, DTS/DAS fibre-based condition monitoring, digital substations (process bus), and participation in ENTSO-E’s RDIC. Digitalisation is also a named förflyttningsområde in Svk’s Strategi mot 2030 (“accelererad digitalisering”).

An early, aspirational parallel on the planning side: in October 2026 the Linux Foundation launched OpenGrid, an open source initiative for a shared data and interoperability layer (open standards, APIs, a data hub and a translator between commercial and open planning tools, first release planned for 2027) aimed at grid planning worldwide; no Swedish or Nordic organisation is named in the announcement (Source - Linux Foundation OpenGrid Launch (2026)).

AI adoption — forecasting, digital twins, and readiness risks

An LTH master’s thesis (Persson, June 2026, with Energiforsk; 17 interviews, 14 of them with DSOs, Svk, RISE and vendors and 3 scoping interviews with Energiforsk staff) gives the most detailed current picture of where AI actually sits in the Swedish energy system, beyond Svk’s own stated priorities above.

Forecasting is where AI has had the largest measured impact so far — not because the algorithms are novel, but because they handle far more data, far more frequently, than manual/traditional methods, shifting parts of grid operation from reactive to proactive. One large DSO has substantially improved system-level consumption forecasts using ML on spot prices, solar irradiance, temperature and historical use, with spot price identified as the single strongest predictor, especially in heavy-industry areas during downturns. A separate, earlier-stage effort — RISE’s PREDATOR/DESGRID project, run with 8 DSOs (Ellevio, Göteborg Energi, Mölndal Energi, Trollhättan Energi, Jönköping Energi, Umeå Energi, C+ Energi, Karlstads Kommun) — found that many Swedish DSOs, especially smaller ones, lack any internal data-driven load-forecasting infrastructure at all to integrate new household-technology-adoption forecasts (EVs, heat pumps, solar) into, the gap the follow-up DESGRID project aims to fill for 10-year network planning. (Source - Energiforsk 2026-1168 AI-modeller Prognostisering Efterfrågan El (2026))

Digital twins in Sweden today are small and single-purpose, not the large interconnected systems described in international literature — e.g. one company’s digital twin of relay protection within a customer’s grid area, another’s twin of building heat demand for district-heating property owners. Svenska kraftnät‘s Kraftsystemhubben has digital-twin-like functionality but its status is debated even internally (see Svenska kraftnät). RISE’s separately funded national grid digital twin, by contrast, has moved well past that early stage: following a 2025 feasibility study, Vinnova funded a 25-organisation consortium (RISE leading; core partners KTH, Svk, Vattenfall, Ellevio, Energiforsk, Power Circle) with ~52 MSEK specifically for the digital twin, inside a larger 320 MSEK “Elflexibel industri” programme running to spring 2029 (initial 3-year phase, possible 3-year extension). Intended uses span multi-vendor communication testing, relay-protection testing across thousands of scenarios weekly, charging-station/solar component testing, and black-start simulation — deliberately built as “a sufficiently realistic model environment” rather than an exact replica, for security reasons. (Source - RISE Digital Tvilling Elnät Elflexibel Industri (2026)) Interviewees converge on two success factors for digital twins generally: design around one clearly defined use case (open-ended “build a digital twin” projects consistently fail), and build bottom-up from the smallest relevant components while agreeing common data standards early enough to allow later interconnection.

Readiness risks the thesis identifies, ranked by the author as the largest barrier to AI adoption overall — competence, not technology maturity or economics: a widening gap between available skill and what the sector needs, worsened by organisational over-trust in AI, engineers without AI/IT grounding being handed fast-rolled-out tools, and a risk that automating “simple” tasks removes the junior staff who would otherwise have caught an AI’s mistakes. This burden falls hardest on small and mid-sized companies, risking competence — and with it, market power — concentrating at large incumbents. Explainability is a related, recurring theme: an unexplainable (“black box”) model is both unaccountable when wrong and harder to audit for tampering. A further named risk specific to this sector: dependency on foreign-hosted cloud infrastructure and firmware-updatable hardware — explicitly including V2G-connected EVs and grid-tied inverters — as a resilience exposure in a state-conflict scenario. (Source - Digitala Monster — AI och Svenska Energisystemet (2026))

Concrete example — Svk’s first digital substation (Åker, 2026)

Svk’s digitalisation strategy names AI/ML forecasting, decision-support automation, and cyber-physical monitoring as R&D priorities (above), but until July 2026 the wiki had no built example. Station Åker (Strängnäs kommun), commissioned 28 July 2026, is Sweden’s first fully digital transmission substation — most control functions run in networked software rather than discrete hardware, cutting copper cabling and enabling earlier pre-installation testing, shorter build time, and remote fault detection. Framed by Svk as necessary to sustain its historiska utbyggnad (historic buildout pace) of the transmission grid at a resource cost it can actually afford per station. Two more digital stations are already planned (Ockelbo, Lindbacka). (Source - Svk Digital Substation Åker (2026))

Concrete example — Vattenfall Eldistribution’s self-healing MV network (2026)

Where Åker (above) is a TSO-side example, Vattenfall Eldistribution’s self-healing grid rollout is the DSO-side counterpart the “Monitoring & control” row otherwise only names in the abstract. Fault indicators at MV stations feed a control-room decision-support system that proposes a network reconfiguration after a fault; the operator reviews and executes it. In a worked example from the release, a ~1,000-customer outage was narrowed to its likely location by two of three fault indicators on the affected line, and a control-room reconfiguration restored ~700 customers within about 3 minutes and a further ~100 from the opposite direction — full restoration, after a field crew located the underlying cable damage and reconnected manually, took about 90 minutes.

As of the release, ~1,200 fault indicators are installed across Vattenfall’s roughly 3,000 MV lines, with the full self-healing concept (indicators plus remote switching plus decision software, working together) live on 200 of them; rollout tracks the ~3%/year rate at which the network is physically rebuilt. The system is decision-support, not autonomous — full automation without an operator in the loop is a stated goal, not current practice. (Source - Vattenfall Eldistribution Självläkande Nät (2026))

Substation digitalization baseline — Ei SGI 2024

Ei’s SGI mandatory reporting (EIFS 2022:5) provides system-wide averages for the digitalization of lokalnät substations across 111 REL reporters in 2024:

Only 35% of lokalnät substations have EIFS 2022:5-compliant hourly measurement. Automation and remote control rates are very low: under 6% for any form of remote operability, under 2% for automated control. This establishes the 2024 baseline — the Swedish distribution grid’s digital transformation remains in early stages for the majority of the network, consistent with DSOs’ generally uneven smart-grid adoption documented above. (Source - Ei SGI Data 2023-2024)

Nätstation maturity — bronze, silver, gold, platinum

A 2018 DNV GL study for Energiforsk’s reference group (Vattenfall, Ellevio, Svk, Kraftringen, Öresundskraft, Karlstad) names the core problem behind the substation statistics above: most Swedish DSOs completely lack information about the relationship between outgoing bays in the fördelningsstation and customer connection points — not a shortage of primary equipment (transformers, cables), but a shortage of information about how the network in between actually behaves. A Vattenfall field trial measuring 14 nätstationer illustrates the scale of the problem directly: only 8 behaved as expected; 1 showed negative losses (documented topology is wrong); 5 showed inexplicably high losses (component faults, bad customer metering, or wrong topology). Over half the trial revealed a data-integrity problem the DSO didn’t know it had, once actual measurement was compared against the assumed network model.

The report proposes a four-tier maturity model for what a given nätstation can do:

The report judges development potential to lie mainly in metering and communication systems, not primary apparatus — and separately notes that ~70% of Swedish local/regional grid components are older than 20 years, ~37% older than 38, unevenly distributed after a large buildout 40–50 years ago, which is why retrofit conditions vary so much station to station. (Source - Energiforsk 2018-540 Framtidens Nätstation (2018))

What DSOs actually do with their data — the 2022 DigiGrid survey

A 34-respondent survey of Swedish DSOs (Power Circle for Energiforsk, summer 2022) puts numbers on a claim otherwise easy to assert without evidence: 85% of respondents said they collect data they do not currently use. When asked what would need to be in place to actually use it, most answers pointed to system support for visualization and analysis — not new data collection. Separately, 56% said there is data they would like to collect but currently cannot, citing insufficient systems/communication pathways, old equipment, IT security, and lack of knowledge and time; one respondent made the cost/value tradeoff explicit: there is an opportunity to collect more from the new smart meters, but currently no value in collecting all of it relative to the cost of validating and storing it.

Among the resources the survey respondents name as critical (competence, personnel, system support, time and money), the competence emphasis falls above all on combined electrical-power and IT/data competence, not either alone — one respondent’s framing was that traditional power-system competence with some programming ability beats deep digitalization competence without enough power-system grounding. This is the same competence-gap conclusion the AI-adoption section above reaches from a different (2026, AI-specific) angle.

A concrete example of the pipeline this section otherwise describes in the abstract: Ellevio built its own load forecasting tool from SCADA data on transmission-grid power draw, combined with producer/large-consumer forecasts and weather, then integrated that tool via API with both SWITCH and NODES to purchase flexibility on sthlmflex — including data-classification work to keep individual consumer information from being exposed. (Source - Energiforsk 2022-895 DigiGrid (2022))

A second, monetizable use case for richer metering — separate from forecasting or flexibility procurement — is non-technical loss (mis-documented networks, theft, other unmetered consumption) localization: a companion RISE study for the same Energiforsk programme found 54 of 159 Swedish network companies had annual losses above 4%, and that cutting those losses to 4% would save an illustrative ~72 million SEK/year (its own rough calculation, at an assumed 50 öre/kWh). The same study states that storage cost for additional metering parameters and resolution is not the major cost driver, compared to the cost of the meters themselves. (Source - Energiforsk 2018-537 Dataanalys och Avancerade Algoritmer (2018))

A vendor-side view of ADMS maturity

AFRY’s whitepaper frames the ADMS as the platform on which DER orchestration, flexibility procurement and the network development plan are run, with a five-level maturity scale from SCADA plus manual outage handling to autonomous self-healing. It cites the JRC DSO Observatory 2024 for large European DSOs: 95 % run a SCADA but only 22 % of MV/LV substations are remotely controllable and only 15 % of EU countries reward flexibility. The paper is consultancy marketing with unsourced benefit figures (Source - AFRY Grid Intelligence Imperative ADMS Whitepaper (2026)).

Why it matters for flexibility

Flexibility is only procurable to the extent the grid is observable and controllable. Without interval metering, baselines cannot be computed (see Baseline Methods); without state estimation, a DSO cannot locate a constraint precisely enough to call a local market; without standardised data exchange, aggregator signals cannot reach the right resources. The digital layer is therefore the enabling substrate for the entire flexibility agenda — and its gaps (fragmented protocols, the not-yet-built DHV/FIS) are among the binding constraints on Swedish flexibility deployment.

Security dimension

A digitalised, flexible grid is also a larger attack surface. The resilience and cybersecurity implications are treated separately in Security and Resilience of the Digitalized Flexible Grid.

Data gaps

• Trends across editions of Ei’s smart-grid report — only the first (R2026:02) exists; the indicator series needs more years to be interpretable
• Whether/how Ei’s R2026:02 indicator set (13 indicators, 3 categories) maps onto or will be revised to match CEER’s 6-dimension, 14-indicator framework — not compared in detail yet
• Quantified SAIDI/SAIFI improvement from Vattenfall’s self-healing MV rollout at network-wide scale (only a single worked example is documented so far, not an aggregate figure)
• Whether any Swedish district-heating company has begun or piloted ancillary-service (stödtjänst) market participation, AI-assisted or otherwise — a 2026 interview study found district-heating firms active only on day-ahead/intraday markets, with no identified ancillary-service activity or AI use for that purpose, despite plausible flexibility resources (CHP, large heat pumps, accumulator tanks) (Source - Digitala Monster — AI och Svenska Energisystemet (2026))