Blogs

What Is an “Intraday Portfolio Signal” [and Why It Matters for Intraday Energy Forecasting]

Written by Melis Karabulut | Apr 20, 2026 12:38:17 PM

Energy Market Insight · Billing and Operations 

Every minute between gate closure and delivery carries a P&L consequence

 

In the Dutch intraday power market, and increasingly across every liberalised European power market, the window between gate closure and delivery is where balance responsible parties (BRPs) and energy suppliers either defend their margins or quietly lose them.

Rooftop PV keeps deepening, heat pumps and EVs keep electrifying the residential fleet, and the “typical” household consumption curve keeps getting less typical.

For any BRP serving millions of customers, forecast accuracy inside the delivery day increasingly determines whether the imbalance bill is an acceptable cost of doing business or a structural margin leak.

This is where the intraday portfolio signal comes in: a real-time, aggregated indicator of how a B2C (or B2B) portfolio is actually behaving right now, compared to how it was expected to behave.

What is an intraday portfolio signal?

An intraday portfolio signal is a near-real-time measurement of actual-versus-expected portfolio behaviour, refreshed continuously throughout the delivery day.

It fuses the latest smart-meter reads, distribution-grid telemetry, and derived production and consumption indicators into a single dynamic variable that tells our forecasting models how the portfolio is deviating from plan, and in which direction.

Unlike a day-ahead forecast, which bakes its assumptions into the schedule and waits for the outturn, an intraday portfolio signal is designed to be used inside the delivery day, during the window where a BRP can still act: re-nominating, trading on continuous intraday markets, or dispatching flexibility before imbalance settles.

In practical terms, it is the layer between raw telemetry and actionable forecasting, the part of modern energy portfolio management that turns live data into real-time steering.

Why historical and weather-only models fall short

Traditional short-term load and PV models lean on two things: historical load curves conditioned on calendar and temperature, and numerical weather predictions translated into irradiance and wind.

Both are well understood, and both perform reasonably well on the median day. The problem is that the median day is no longer where most of the risk lives.

Human behaviour does not respect climatology. A sudden cold front, an unexpected school day, a broadcast event, a rapid cloud transit over the Randstad: each of these can push residential demand or distributed PV well outside what any historical-plus-weather model would predict for that quarter-hour.

As residential electrification deepens and PV penetration grows, the tails of the distribution get fatter. A forecasting stack that relies exclusively on ex-ante inputs is, by construction, blind to what is actually happening in the portfolio between gate closure and delivery.

What actually goes into the signal

The intraday portfolio signal is not a single data stream. It is an aggregation and normalisation layer that reconciles several asynchronous, imperfect inputs into a coherent live view of the portfolio:

  • Near-real-time smart-meter telemetry from a representative sample of the portfolio, aggregated to quarter-hourly granularity.
  • Grid-level measurements (DSO and TSO feeds) that capture net system flows and let us sanity-check sampled meter data against system reality.
  • Solar nowcasting from satellite-derived irradiance and sky-imager data, compared against the PV forecast curve for our fleet.
  • Portfolio state (churn, new connections, heat-pump and EV activations), so the denominator the signal is normalised against reflects today’s portfolio, not last quarter’s.

The signal itself is not the raw data; it is the residual: the delta between what the portfolio is doing and what we forecast it would do for this quarter-hour, expressed in a form that downstream models can consume directly. Latency matters. We engineer for minutes, not hours.

From signal to decision: forecasting, trading, and risk management

Inside our intraday stack, the signal plays three roles simultaneously.

  • Model correction. Forecasting models consume the signal as a real-time bias term. If the residual has been negative for several quarter-hours, the next-hour forecast adjusts immediately rather than waiting for the next retrain cycle.
  • Regime detection. Sharp excursions are often the earliest indicator that the market is in a different regime than day-ahead assumed: a faster PV ramp, an unexpected demand surge on a heat-pump-heavy cohort, or a customer segment behaving out of character.
  • Trading and re-nomination input. The same signal that corrects the forecast sharpens short-term volume expectations at the continuous intraday window, tightening the gap between expected and contracted position before imbalance kicks in.

The practical effect is that forecasting stops being a “set at 11:00 D-1 and hope” exercise and becomes a continuously calibrated process, closer to how modern energy trading and risk-management desks already think about position management.

Why this matters for balancing exposure and margins

The Dutch imbalance mechanism is unforgiving by design. Under TenneT’s single-price settlement, systematic forecast errors are monetised in real time at prices that routinely diverge sharply from day-ahead. For a supplier with material B2C volume, even a persistent one-to-two-percent systematic bias on residential load can translate into a seven-figure annual drag, before considering the amplified impact on volatile days.

A robust intraday portfolio signal helps on four concrete fronts:

  • Reduces balancing risk by detecting and correcting forecast drift early in the delivery day, before it compounds across multiple quarter-hours.
  • Accelerates reaction time to surges or drops in household consumption or PV output, whether weather-driven or behavioural.
  • Sharpens short-term trading decisions at the continuous intraday window, improving trade sizing and timing.
  • Stabilises margins in a market where renewables penetration keeps widening the distribution of possible outcomes.

The value is not in having more data. It is in transforming live data into actionable signals that directly influence portfolio decisions, which is increasingly what energy portfolio management and energy risk management look like in practice.

Better signals. Better forecasts. Better outcomes. 

The Dutch system’s trajectory toward a more weather-driven, more decentralised, more flexibility-rich grid is no longer a projection; it is already well in motion, and every liberalised European market is moving along the same curve. Every incremental percentage point of residential PV, every thousand heat pumps commissioned, every EV charger that joins the fleet widens the intraday uncertainty band.

BRPs and suppliers that rely solely on ex-ante models will feel this as rising imbalance costs and compressed margins. Operators that invest in real-time steering, ingesting their own portfolio’s behaviour as a first-class input to the forecast, will increasingly separate themselves by operating with tighter error bands and lower balancing exposure.

The goal is no longer just to forecast the future. It is to continuously steer toward it with the best information available.

At Eneve, that is the flow we are built to go for.

Explore how Eneve works here.