FlexSource - BeFlexible D5.1 Demo Planning and Deployment (2024)

Source - BeFlexible D5.1 Demo Planning and Deployment (2024)


BeFlexible project deliverable D5.1 — “Report on Demo Planning and Deployment.” Final version, submitted 2024-08-30. Part of WP5 (North EU Demo — Long Track), covering the planning and first season (V2023/24) of DEMO 2, the long-track demonstration following DEMO 1 (the fast-track CoordiNet demonstration).

Lead author organization: E.ON Energidistribution. Contributors: E.ON, E.ON EIS (Energy Infrastructure Services), SIN (Social Innovation Norway), RWTH Aachen University. EU-funded (Horizon Europe).

Raw file: Raw/PDF extractions/BeFlexible-D5.1-Report-on-demo-planning-and-deployment-1/


Document scope

D5.1 covers five areas:

  1. Market environment and customer research for FSP recruitment in Sweden
  2. Value propositions and business models for flexibility
  3. FSP engagement and recruitment methodology and V2023/24 experience
  4. Local flexibility market and product design — V2023/24 state and V2024/25 proposals
  5. SWITCH platform development — architecture, challenges, and redesign rationale
  6. E.ON EIS ectocloud integration for aggregated BESS/heat pump flexibility
  7. Congestion forecast improvement — RWTH probabilistic RNN and E.ON TFT models

Key claims

Market context (Sweden, 2024)

  • Swedish electricity consumption projected to double to 300–400 TWh; Svenska kraftnät committed to adding 2,500–3,000 MW in southern Sweden by 2027 to meet LOLE target (1 hour/year)
  • Ancillary market turnover projected to reach ~SEK 4 billion in 2024, with high prices attracting battery FSPs and new software entrants
  • Home batteries in Sweden grew 15,000 → 40,000 units between 2022 and 2023
  • V2G potential: Ei + PowerCircle estimated ~50 GWh battery capacity available from EVs by 2035

Customer segments for FSP recruitment

Qualitative research identified six target segments: property owners, property developers, warehouse/logistics, industrial companies, smaller property owners, municipalities.

Master thesis (Linköpings University, tutored under BeFlexible) on energy-intensive industry barriers:

Top barriers (ranked by industries):

  1. Hidden costs (including lost production)
  2. Sensitive processes (cannot interrupt at any moment)
  3. Interconnected processes

Top drivers:

  1. Compensation/reimbursement
  2. Process controllability
  3. Political signals

Key finding: many industries expressed interest but lacked knowledge of compensation levels and prequalification requirements — implying a communication gap, not a fundamental objection.

V2023/24 flexibility markets (first BeFlexible demo season)

Three markets ran under BeFlexible in V2023/24:

MarketRequirementDurationMin resourceProcurementPlatformSeason
Hässleholm2 MW90 h0.1 MWDirect procurementSWITCHNov 2023–Mar 2024
Southern Skåne10 MW100 h0.1 MWDirect procurementSWITCHNov 2023–Mar 2024
Stockholm (sub-areas)0.5–1 MW/market80–160 h0.1 MWDirect procurementNODESDec 2023–Mar 2024

Compensation threshold in V2023/24: Availability compensation was forfeited if less than 50% of promised flexibility was delivered. (Changed to 75% in subsequent seasons.)

Direct procurement limit: SEK 1,200,000 per market per product — the binding constraint that triggered the DIS (Dynamiskt Inköpssystem) development.

E.ON EIS Sjölunda heat pump — first delivery data

E.ON EIS’s Sjölunda district heating heat pumps (Malmö) participated in the Södra Skåne Flexibility Market in V2023/24 as a pilot for integrating district heating assets:

  • Available to market: 2.5 MW potential downregulation per occasion
  • Availability orders received from Skåne flex market: 8 occasions
  • Total available volume: 152 MWh
  • Actual deliveries: 6 activations / 57 MWh delivered

The Sjölunda pilot demonstrated district heating coupled assets as viable DSO flexibility providers and initiated work to scale to more E.ON EIS assets. Capabilities were being developed for V2024/25 integration via ectocloud.

Market longevity commitments (as of August 2024)

MarketStated active until
Bålsta≥ 2029
Enköping≥ 2029
Hässleholm≥ 2028
Kallhäll≥ 2027
Kungsängen≥ 2027
Nordöstra Skåne≥ 2028
Norra Örebro≥ 2030
Södra Skåne≥ 2029
Vaxholm≥ 2027

Product design evolution — V2023/24 problems and V2024/25 proposals

The V2023/24 products were named “Availability Orders”, “Seasonal Availability”, and “Free Bids” — the precursors to today’s TO, ST, and DO abbreviations.

Availability Orders (→ Tillgänglighetsordrar / TO):

  • V2023/24 issue: “first-come, first-served” selection — found non-compliant with Swedish public procurement law (LOU/LUF)
  • V2024/25 proposal: replace with “lowest-bid-wins” auction
  • Trading time collision: existing clearing times overlapped with other markets, forcing FSPs to choose
  • “Guaranteed compensation” (Garantiersättning) term caused FSP confusion — proposed for removal

Seasonal Availability (→ Säsongstillgänglighet / ST):

  • V2023/24 issue: 4-hour endurance requirement locked batteries out of FCR-D during committed blocks
  • V2024/25 proposal: shorter, individually configurable endurance per market period; endurance-weighted bid selection

Free Bids (→ Direktordrar / DO):

  • V2023/24 issue: FSPs submitted bids before DSO published need — non-compliant with procurement rules (“bidding blindly”)
  • V2024/25 proposal: DSO publishes need first; FSPs then have a trading window
    • DA window: D-2 18:30 → D-1 08:30, clearing D-1 09:00
    • ID window: D-1 15:00 → H-4, clearing H-3

DIS (Dynamiskt Inköpssystem):

  • Alternative to direct procurement (<SEK 1.2M) or full public tender
  • Investigation ongoing as of Aug 2024; implementation uncertain before V2024/25 start
  • Enables continuous FSP recruitment within a pre-qualified supplier framework

Product naming:

  • Renaming to Svk-style abbreviations (ST/TO/DO) was under discussion, to be finalized during 2024

SWITCH platform — pre-redesign state and challenges (early 2024)

Platform data model and architecture unchanged since CoordiNet 2022. A full platform redesign was initiated in early 2024 for launch before V2024/25. D5.1 describes the then-current (pre-redesign) state.

Technical stack: Azure cloud, .NET/C# backend, REST APIs, Azure AD authentication.

Challenges documented:

Baseline tampering (2022/23): During V2022/23, a risk was identified of FSPs altering their register configuration to manipulate metering data. Fix: resource control moved to DSO — FSPs submit qualification requests; DSO approves and owns the resource record. FSP can view but not modify. Separate baseline upload endpoint added with stricter validation (no changes after a trade has occurred).

Regulatory moving target: NC DR and national procurement rules (LOU/LUF) imposed requirements on bid recommendation algorithms mid-development. Platform requirements had to be kept flexible for not-yet-finalized regulation.

Integration barriers: Support for availability-based products made full FSP API integration more complex. Simplified API endpoints planned for V2024/25. Onboarding documentation and training sessions partially bridged the gap.

Business/development misalignment: Structured feedback loops between business stakeholders and software developers identified as critical. Product requirements need to originate from operational stakeholder needs, not internal technical assumptions.

SWITCH forecasting — E.ON TFT model

During 2023, E.ON developed a new load forecasting engine based on Temporal Fusion Transformers (TFT) — a Google Research architecture combining RNNs, transformers, and attention mechanisms for multi-horizon time series forecasting.

Inputs: years of historical power load meter readings + weather data (actuals and forecasts)

Output: predicted hourly load for the next 48 hours; updated each hour. Weekly forecasts for prioritized grid points.

Infrastructure: PyTorch training on GPU; hosted in Microsoft Azure ML Studio; integrated with E.ON’s data lake.

Application: predicts load at congested substations (various voltage levels). Quantifies congestion events with sufficient precision to automate the full flexibility market data flow — from need calculation through market clearing to settlement.

Replaced the earlier GRU-type RNN as more scalable. RWTH (project partner) developed a parallel probabilistic RNN model using encoder-decoder architecture and quantile regression, for research comparison purposes.

E.ON EIS ectocloud integration

ectocloud™ is E.ON EIS’s IoT platform for behind-the-meter energy management. D5.1 describes development of a SWITCH-ectocloud integration to enable aggregated flexibility delivery from BESS and heat pumps to local flexibility markets.

Four software agents being developed:

  • Broker — communicates with SWITCH API, handles order confirmation
  • Resource Scheduler — verifies resource availability, books operation schedule
  • Forecaster — predicts asset capability
  • Executer — manages delivery day dispatch per schedule

Nine potential pilot sites identified (batteries and heat pumps). Target: automated delivery of >100 kW aggregated from multiple assets by V2024/25.

Villkorade avtal as fallback in the flexibility stack

D5.1 provides explicit description of how Villkorade Avtal and flexibility markets interact:

  1. A customer with a villkorat avtal is controlled as a last resort — only if market-based flexibility is insufficient
  2. A customer with a villkorat avtal who also participates in the Flexibility Market can “jump the queue” — their bid competes on the market like any other; if it wins, they are compensated normally
  3. The bid does not need to match the conditional capacity in their villkorat avtal agreement exactly
  4. Result: villkorade avtal customers increase FSP pool and competition, lowering costs for the DSO

Relevance to existing wiki topics

TopicNew information
SWITCHTFT forecasting model; baseline tampering incident and fix; pre-redesign state; ectocloud integration architecture
Flexibility MarketV2023/24 market parameters; market longevity table; old 50% delivery threshold; Sjölunda delivery data
Villkorade AvtalQueue-jumping mechanism (villkorat avtal holder can bid into flex market and “jump the queue”)
E.ON EnergidistributionEIS asset flexibility; ectocloud integration; FSP engagement methodology
Demand ResponseMaster thesis findings on industrial barriers/drivers
Network Code on Demand ResponseRegulatory moving-target effect on SWITCH product design confirmed