Reserves: what catches the gap

The day-ahead plan is always a little wrong. Wind blows harder than forecast. A factory turns on a furnace. A nuclear unit trips. The plan and reality drift apart, and the system has to close the gap in real time.

It does this with three layers of reserves. Each one is slower and deeper than the one before. The fire department gives a good picture for this.

• A smoke alarm goes off first. Fast, automatic, short-lived.
• A kitchen extinguisher comes next. Slower, but bigger.
• A fire engine arrives last. Slowest, but it actually puts the fire out.

The grid works the same way.

The three layers

flowchart TB
    EV([0 seconds<br/>generator trips, or demand spikes])

    EV --> INE([Inertia<br/>spinning rotors slow the fall<br/>buys a few seconds<br/>free])

    INE --> FCR([FCR<br/>units push more power in automatically<br/>STOPS the fall<br/>reacts in seconds])

    FCR --> AFRR([aFRR<br/>signal from Svenska kraftnät<br/>PULLS frequency back to 50 Hz<br/>reacts in minutes])

    AFRR --> MFRR([mFRR<br/>operator activates bids<br/>FREES UP aFRR for the next event<br/>15 minute activation])

    style EV fill:#fecaca,stroke:#b91c1c,color:#7f1d1d
    style INE fill:#fef3c7,stroke:#a16207,color:#713f12
    style FCR fill:#fed7aa,stroke:#c2410c,color:#7c2d12
    style AFRR fill:#fed7aa,stroke:#c2410c,color:#7c2d12
    style MFRR fill:#fed7aa,stroke:#c2410c,color:#7c2d12

Three verbs, in this order: contain, restore, replace. If you remember those, the rest is just naming.

A real example, step by step

A 600 MW unit at Ringhals trips offline at 14:23 on a Tuesday. Demand does not change. Suddenly there are 600 MW more demand than supply.

sequenceDiagram
    autonumber
    participant F as Frequency
    participant I as Inertia
    participant C as FCR-D
    participant A as aFRR
    participant M as mFRR

    Note over F: 14:23:00, Ringhals trips
    F->>I: frequency falls fast
    I-->>F: slows the fall for ~2 seconds
    F->>C: frequency reaches 49.85 Hz
    C-->>F: FCR-D activates, fall stops
    Note over F: 14:23:30, stable but low
    F->>A: TSO signals aFRR to increase
    A-->>F: frequency climbs back to 50 Hz
    Note over F: 14:30, back at 50 Hz
    A->>M: TSO activates mFRR bids
    M-->>A: mFRR takes over
    Note over F: aFRR is free again

The whole event takes 15 to 30 minutes. Nobody in Stockholm noticed their lights flicker. But somewhere a battery in Skellefteå dumped power for 90 seconds. Somewhere a hydro plant in Jämtland opened a valve a little more. The market paid those assets to be ready, and is now paying them for the energy they delivered.

You are mostly paying for readiness, not energy

This is the part that catches people new to the industry.

A reserve product mostly pays a fee for being available, in SEK per MW per hour. A battery sitting at half charge, doing nothing, still gets paid. The energy delivered when an event finally happens is a smaller bonus on top.

Why? Because the system cannot wait until something goes wrong before finding a unit. The unit has to be armed and contracted in advance. The capacity fee buys that readiness.

Why this is where the real work lives

If you join a Swedish energy company today and your job touches dispatch, trading algorithms, or asset optimisation, you will spend most of your time on these reserve markets. Not on day-ahead.

Day-ahead is a screen with a number on it. The reserve markets are where the operational complexity lives. Tight timescales. Noisy data. Real automation. Real revenue. Batteries earn their best returns here, in FCR-D.

Next

That covers the system half. Now the physical half: how electricity actually travels from a generator to your wall. See From generator to your socket.