Generate a standard load profile for gas • standardlastprofile

A standard load profile reflects the typical pattern of gas consumption. It is used for forecasting and balancing purposes for various customer types or customer groups for which no continuous measurement data is available. In this respect, standard load profiles represent a simplification.

In this article we explain how slp_gas() implements the algorithm of the synthetic procedure (German: synthetisches Verfahren), following the definitions by BDEW, VKU, and GEODE in the Leitfaden Abwicklung von Standardlastprofilen Gas (2025).¹

Synthetic procedure

The synthetic procedure takes a bottom-up approach: the gas consumption for a gas network is explained by the sum of the individual gas consumers within that network.²

For each customer type, i.e. profile, the gas consumption of day \(D\) is the result of three factors:

outside temperature
preferences and habits
the day of the week

Mathematically, this is expressed as:³

\[ Q(D) = h(\vartheta_D) \times KW \times F_{WT, D} \] where:

\(h(\ldots)\) is a profile function evaluated at the daily temperature \(\vartheta_D\).
\(KW\) is a customer-specific scaling factor in kWh/day (German: “Kundenwert”).
\(F_{WT}\) is a weekday factor that adjusts for consumption differences across days of the week.

Profile function \(h(\vartheta_D)\)

Gas consumption is strongly correlated with outdoor temperature; the colder it gets, the more gas is consumed. The relationship between outside temperature and gas consumption is captured by the profile function \(h(\vartheta)\).

This function was initially published in 2003 as an S-shaped sigmoid function:⁴

\[ h(\vartheta) = \frac{A}{1 + \left(\dfrac{B}{\vartheta - \vartheta_0}\right)^C} + D \]

Since then, it has been revised as part of an ongoing debate. In a research paper from 2015 it was found that a pure sigmoid approach systematically underestimated consumption at very cold temperatures. The authors proposed an improved formulation called SigLinDe (Sigmoid + Linear + Deutschland).⁵ SigLinDe adds a linear component to that sigmoid equation representing space-heating demand (German: Heizgas-Gerade) and hot water demand (German: Warmwasser-Gerade).

\[ h(\vartheta) = \underbrace{\frac{A}{1 + \left(\dfrac{B}{\vartheta - \vartheta_0}\right)^C} + D}_{\text{sigmoid part}} + \underbrace{\max\!\left(m_H \vartheta + b_H,\; m_W \vartheta + b_W\right)}_{\text{linear part}} \]

The SigLinDe approach was adopted by BDEW and has since been the binding standard.⁶ slp_gas() implements the SigLinDe approach. The parameters \(A\), \(B\), \(C\), \(D\), \(\vartheta_0\), \(m_H\), \(b_H\), \(m_W\), \(b_W\) are profile-specific constants.⁷

All 15 gas profile IDs share this one SigLinDe function — they differ only in their parameter values (\(A\), \(B\), …, \(b_W\)). The complete set is tabulated in the SigLinDe Parameters and Weekday Factors article and returned in R by slp_gas_coefficients():

slp_gas_coefficients("HEF")
#>   profile_id variant        A         B        C         D theta0         mH
#> 1        HEF      34 1.381966 -37.41242 6.172318 0.0396284     40 -0.0672159
#> 2        HEF      33 1.620954 -37.18331 5.672785 0.0716431     40 -0.0495700
#>          bH         mW        bW
#> 1 1.1167138 -0.0019982 0.1355070
#> 2 0.8401015 -0.0022090 0.1074468

The HKO profile (cooking and hot water only) is a special case: all four linear coefficients (\(m_H\), \(b_H\), \(m_W\), \(b_W\)) are zero, so the linear term \(\max(m_H\vartheta + b_H,\; m_W\vartheta + b_W)\) vanishes and \(h(\vartheta)\) collapses to the bare sigmoid — it is the original pre-SigLinDe TU München profile, carried over unchanged. Because the variant 33 / 34 split only affects the linear part, the two variants are identical for HKO.

The result of \(h(\vartheta)\) is a number > 0 without a dimension:

# profile "GMF", Ausprägung: 34
slp_gas_siglinde(
  theta  =   8.0,
  A      =   1.0443538,
  B      = -35.0333754,
  C      =   6.2240634,
  D      =   0.0502917,
  theta0 =  40.0,
  mH     =  -0.053583,
  bH     =   0.9995901,
  mW     =  -0.0021758,
  bW     =   0.1633299
)
#> [1] 1

The result above is exactly 1 — which is no coincidence. The SigLinDe coefficients of every profile are calibrated so that the profile function equals 1 at the reference temperature of 8°C. At 8°C the profile term drops out of \(Q(D) = KW \times h(\vartheta_D) \times F_{WT,D}\), so the customer value \(KW\) is simply the daily consumption on an 8°C day — which is exactly how it is defined.

HKO is the one exception, and precisely because of its missing linear part. The SigLinDe profiles owe that exact-1 value at 8°C to the linear term, which is fitted (together with the sigmoid) to pin \(h(8\,\text{°C}) = 1\). HKO has no linear part, so its bare sigmoid is not pinned to the reference and reaches about 1.06 at 8°C instead — visible in the BDEW profile sheet, where the HKO row lists \(h(8\,\text{°C}) = 1.05612\).

The temperature \(\vartheta\) fed to this function is the daily allocation temperature (defined next); the dimensionless \(h(\vartheta)\) is later scaled to kWh by the customer value \(KW\) (covered below).

Allocation temperature

The allocation temperature \(\vartheta_D\) is the daily temperature fed to the profile function \(h(\vartheta)\). Two ways to compute it are defined in the Leitfaden:⁸

Simple daily mean — the arithmetic mean of the 24 hourly values over the gas day:

\[\vartheta_D = \frac{1}{24} \sum_{i=1}^{24} T_i\]

Geometrically-weighted 4-day mean — recommended by BDEW for grid operators, giving more weight to the current day:

\[\vartheta_D = \frac{T_D + 0.5 \cdot T_{D-1} + 0.25 \cdot T_{D-2} + 0.125 \cdot T_{D-3}}{1 + 0.5 + 0.25 + 0.125}\]

slp_gas() accepts whichever values you pass in temperatures — the choice of method is the caller’s responsibility. If you have raw daily means from DWD and want the geometrically-weighted variant, apply it before calling slp_gas(). (If you already have gas forecast temperature (GPT) values from a meteorological service provider, use those directly — GPT is not this formula.)

# Geometrically-weighted 4-day mean (more weight on the current day):
weights <- c(1, 0.5, 0.25, 0.125)

allocation_temp <- \(temps) {
  as.numeric(
    stats::filter(temps, weights / sum(weights),
                  method = "convolution", sides = 1)  # past values only
  )
}

# Apply to your daily mean series before passing it to slp_gas().
# The first three days come back NA (no 3-day history yet):
# temps_weighted <- allocation_temp(temps)

Kundenwert

The customer value (kundenwert) accounts for a customer’s individual consumption behaviour. It is a scaling factor in kWh/day — specifically, the daily gas consumption at the reference temperature of 8°C, the point at which \(h(\vartheta) = 1\) for every profile. We can rearrange the gas consumption equation above and write \(KW\) as:

\[ KW = \frac{E_a}{\displaystyle\sum_{D} h(\vartheta_D) \times F_{WT,D}} \]

Where

\(E_a\) is the annual gas consumption
\(\vartheta_D\) is the daily allocation temperature for day \(D\)
\(F_{WT,D}\) is the weekday factor for day \(D\)

Using the example of a single-family home (HEF) in Düsseldorf with an annual consumption of 15,000 kWh — the same customer as in the README — the table below shows how the denominator is built up, day by day, from the Düsseldorf long-term daily mean temperatures (DWD station 1078, 2004–2024):

Date	°C	h(ϑ)
Jan 1	4.8	1.3962
⋮	⋮	⋮
Jun 30	19.2	0.1727
Jul 1	18.9	0.1765
Jul 2	19.2	0.1727
⋮	⋮	⋮
Dec 31	5.7	1.2833

The sum in the denominator across all 365 days gives \(\sum_D h(\vartheta_D) \times F_{WT,D}\) = 272.32, so \(KW \approx\) 15,000 / 272.32 = 55.1 kWh/day.

With \(h(\vartheta_D)\) and \(KW\) in hand, the only remaining factor in \(Q(D) = KW \times h(\vartheta_D) \times F_{WT,D}\) is the weekday factor \(F_{WT,D}\).

Weekday factors

Unlike the electricity profiles, which group Monday–Friday into a single workday type, the gas procedure uses a separate scaling factor for every weekday from Monday through Sunday.⁹

Public holidays are treated as a Sunday.¹⁰ 24 December and 31 December are treated as Saturday unless they fall on a Sunday; this convention mirrors the established rule from the BDEW electricity SLP procedure and is applied here by analogy.

For the residential profiles HEF, HMF, and HKO, however, all weekday factors are equal to 1, that is, gas consumption in households is assumed not to differ significantly by day of the week. Commercial profiles on the other hand show visible weekday patterns, the range of the weekday factors is 0.3877 - 1.2707, illustrated by the chart below:

Profile IDs

There are 15 gas profile IDs defined in the BDEW Leitfaden, split into residential and commercial / industrial types.

Residential profiles

Residential profiles share two important characteristics: all weekday factors equal 1 (no day-of-week differentiation), and heating load is the dominant driver of demand variability.

Profile	Description	Notes
`HEF`	Single-family home (Einfamilienhaus)	Strongest seasonal swing; includes space heating + hot water
`HMF`	Multi-family home (Mehrfamilienhaus)	Lower per-unit demand than `HEF` due to shared building envelope
`HKO`	Cooking and hot water only (Kochen / Warmwasser)	No space heating; TUM profile — sigmoid only, linear part is zero

HKO is appropriate for customers whose gas connection serves only a cooker or instantaneous water heater. Its flat, temperature-insensitive shape is visible in the SigLinDe curve plot above.

Commercial and industrial profiles

Commercial profiles have profile-specific weekday factors that reflect the operating hours and schedules of each sector. They are grouped here by consumption pattern:

Process-heat dominated (relatively flat seasonal pattern, strong weekday structure):

Profile	Description
`GWA`	Laundries — large hot-water demand, insensitive to outdoor temperature
`GBA`	Bakeries — continuous baking process, pronounced weekday factors
`GPD`	Paper and printing — industrial process heat

Space-heating dominated (strong winter peak, moderate weekday variation):

Profile	Description
`GKO`	Small commercial (Kleinstgewerbe)
`GHA`	Trade and commerce (Handel)
`GBD`	Services (Dienstleistung)
`GBH`	Accommodation (Beherbergung)
`GGA`	Gastronomy (Gastronomie)
`GMK`	Metal and automotive (Metall / Kfz)
`GGB`	Mixed commercial (gemischtes Gewerbe)
`GMF`	Large multi-family / mixed use (Mehrfamilienhaus groß)

Aggregate / summary profile:

Profile	Description
`GHD`	Trade, commerce and services aggregate (GHD-Stützpunkt) — weighted mean across all G-types

GHD is used when a customer cannot be assigned to a specific sector. Its parameters are a weighted average of the individual commercial profiles and its weekday factors are close to but not equal to 1.

A working example: Düsseldorf

With the algorithm covered, let us see how to use slp_gas() in practice. slp_gas() takes: - a profile_id - a dates vector, and a matching temperatures vector, and - a kundenwert.

kundenwert is the customer-specific scaling factor in kWh/day.

Deriving the Kundenwert

To produce a meaningful result the temperature series should cover a full year at least. You can use measured data from a trustworthy institution like the German Meteorological Service, or DWD for short. The rdwd package provides a straightforward interface to the open-data portal of the DWD (Deutscher Wetterdienst).¹¹ No registration or API key is required.

rdwd package

The workflow has three steps:

selectDWD() — find the download URL for a station and dataset.
dataDWD() — download and unzip the file.
readDWD() — parse the file into a data frame.

For gas SLPs, we need the daily climate summary (var = "kl", res = "daily"). The column TMK contains the daily mean temperature in °C. To find the exact station name to pass to selectDWD(), use findID() or open the interactive station map:

rdwd::findID("Freiburg")             # returns matching station names and IDs
rdwd::selectDWD(mapDWD = TRUE)       # opens an interactive map in the browser

The Kundenwert is derived once, from the customer’s annual consumption and a reference temperature series for their location, using slp_gas_kundenwert(). The resulting value is then passed to slp_gas() for whatever period you want to generate — a single month, a quarter, the whole year.

Keeping the derivation a separate step is deliberate. \(KW\) is a property of the customer and their local climate, so it should be computed from a representative full year — ideally a multi-year mean.

Take the single-family home (HEF) from the README’s example: located in Düsseldorf, with an annual gas consumption of 15,000 kWh. For the reference temperatures we use the long-term daily mean over 2004–2024 — a multi-year (climatological) mean rather than a single year, so that no individual-year anomaly distorts the scaling factor¹².

library(rdwd)

# Full daily climate record for Düsseldorf (DWD station 1078)
links <- selectDWD("Duesseldorf", res = "daily", var = "kl",
                   per = c("historical", "recent"))
raw   <- do.call(rbind, readDWD(dataDWD(links, read = FALSE, force = TRUE),
                                varnames = FALSE))

# Long-term mean daily temperature over 2004-2024, one value per calendar day
ref       <- raw[format(raw$MESS_DATUM, "%Y") %in% as.character(2004:2024), ]
ref$mmdd  <- format(ref$MESS_DATUM, "%m-%d")
clim      <- aggregate(TMK ~ mmdd, data = ref, FUN = mean)
clim      <- clim[clim$mmdd != "02-29", ]          # drop the leap day
dates_ref <- as.Date(paste0("2023-", clim$mmdd))   # any non-leap year
temps_ref <- clim$TMK

# kundenwert for a 15,000 kWh/a HEF customer
slp_gas_kundenwert("HEF", dates_ref, temps_ref, annual_consumption = 15000)
#>      HEF
#> 55.08344

This Düsseldorf customer has a Kundenwert of about 55.1 kWh/day. That fixed value is what the README example passes to slp_gas() to generate the 2025/26 heating season — and it is what we use below to ask how the same customer would fare in a different climate.

What if the customer moves?

We keep the Düsseldorf customer’s Kundenwert of 55.1 kWh/day fixed and ask what they would consume in a colder climate — same customer, same KW, different weather. Germany spans several distinct climate zones; we compare three:

Chemnitz — continental climate in the Erzgebirge foothills of Saxony; colder winters and the greatest temperature amplitude of the three.
Freiburg im Breisgau — mild oceanic-continental climate in the sheltered Upper Rhine Plain; the warmest, sunniest large city in Germany.
Hamburg — maritime climate on the North Sea coast; mild but cloudy, windy winters.

The grid below has one column per comparison city and one row per month of the 2025/26 heating season (the same period as the README example). Each point is a day, plotting the city’s daily gas consumption (y-axis) against Düsseldorf (x-axis); the 45° line marks equal consumption.

That winter all three cities ran colder than Düsseldorf, so every point cloud sits above the line — the question is by how much:

Chemnitz is highest, about 35 % more gas over the season (≈ 16,400 vs 12,200 kWh);
Hamburg follows at about 25 % more;
Freiburg is closest to Düsseldorf at about 11 % more.

The gap is widest on the cold January and February days at the top-right of each panel, and narrows in the mild shoulder months of October and April where the clouds approach the line.