Enerlytics AIVisit Website →
← Back to blog

Weather Normalization for Energy: HDD, CDD and UK Method

November 3, 2026
Weather Normalization for Energy: HDD, CDD and UK Method

Weather normalization restates your energy use as if every period had faced the same weather, so you can tell a real efficiency change from a cold winter or a hot summer. In the UK you do it with degree days: you measure how far outdoor temperature fell below (heating) or rose above (cooling) a base temperature, then relate your consumption to that figure. This guide explains how it works, how HDD and CDD differ, which base temperature to pick, and where the method stops being reliable.

Enerlytics AI
See What Drives Your Energy Use
Enerlytics AI provides half-hourly consumption analytics to help UK businesses track energy use and assess changes across their sites.
Explore energy analytics

TL;DR

  • Weather normalization estimates what a building would have used under average weather, so unusual hot or cold spells do not distort year-on-year comparisons (ENERGY STAR).
  • UK heating degree days (HDD) traditionally use a 15.5°C base; for most UK business buildings, HDD does most of the work and CDD matters mainly where you run air conditioning.
  • The core model is simple: consumption = base load + slope × degree days. Fit it on your own half-hourly or monthly data, then ask whether you used more than the weather justified.
  • Normalization fixes time-series comparisons. It does not make two sites in different climates, or with different opening hours, directly comparable.

Table of Contents

What weather normalization actually does

Energy use in a building with heating or cooling moves with the weather. If this January was colder than last January, your gas consumption will rise even if nothing about the building or how it is run has changed. Comparing raw totals tells you the weather was different. It does not tell you whether your estate got better or worse.

Normalization removes that noise. ENERGY STAR describes the idea as estimating the energy a building would have used under average weather conditions, so that unusually hot or cold periods do not distort the assessment of consumption over time (ENERGY STAR Portfolio Manager technical reference). The US guidance behind that reference is American, but the logic is universal: control for weather and other external drivers so that comparisons are meaningful, and so that operational changes stand out from seasonal ones (Benchmarking to Save Energy).

The same guidance adds a practical point: normalization can use simple or sophisticated statistical techniques, and whichever you choose should be tested to see how well it predicts actual performance. We return to that test below, because it is the step most teams skip.

The question to ask of every bill

A useful way to frame the result, put forward in UK monitoring and targeting training, is to stop asking "was it higher than last year?" and ask whether consumption exceeded the amount justified by conditions (Energy Academy UK). A 6% rise in gas use during a winter that was 10% colder than the year before is not a problem. A 2% fall in a winter that was 10% milder may be one.

Degree days in plain terms

A degree day condenses a day's outdoor temperature into one number that is relevant to a building. CIBSE Technical Memorandum 41 defines degree days as the sum of differences between outdoor temperature and a reference base temperature over a specified period, used to analyze weather-related building energy consumption (CIBSE TM41). Published degree-day figures reduce local or regional weather to a single weekly or monthly number showing how cold, or for cooling how hot, the period was (VESMA guide).

How HDD is calculated

For UK heating degree days, the traditional base is 15.5°C. On a day when the mean outdoor temperature is below 15.5°C, HDD is the base minus that mean. If the mean is above the base, HDD for the day is zero. Daily values are then summed over the month, quarter or year (UK Government analysis of domestic energy bills).

A worked example from UK training material: a day averaging 10.5°C against a 15.5°C base contributes 5 HDD, while a day averaging 18°C contributes none (Energy Academy UK). Add up the days in a month and you have that month's HDD.

Published degree days are calculated from daily maximum and minimum air temperatures, and local continuous measurements can improve accuracy (VESMA guide). For context, a standard UK table lists an annual total of 2,463 heating degree days at a 15.5°C base (VESMA standard degree days). Your own year will land above or below that depending on the weather and region.

HDD vs CDD: which one explains your site?

CDD is the mirror image of HDD. It counts how far temperatures rise above a cooling base, and it indicates air-conditioning load (Energy Academy UK). The UK Met Office equations published in an academic paper use 15.5°C for heating and 22°C for cooling, with formulas that change depending on whether the base falls within the day's temperature range (Nature Communications, Table 1).

HDD (heating)CDD (cooling)
What it countsDegrees the daily mean falls below the baseDegrees the daily mean rises above the base
Typical UK heating base15.5°C (18.5°C or 10°C for some buildings)n/a
Typical cooling basesn/a15.5°C for mainly summer cooling, 5°C for year-round cooling
Main fuel it explainsGas, oil, electric heating, heat pumpsElectricity for chillers and air conditioning
Where it matters in the UKAlmost every occupied buildingOffices with comfort cooling, data rooms, cold stores, retail

The base temperatures in the table come from UK guidance: 15.5°C is the traditional heating base, 18.5°C may suit hospitals, and 10°C may suit modern buildings with substantial internal gains or insulation. For cooling, 15.5°C suits installations cooled mainly in summer, and 5°C is described as an all-purpose base for buildings that need cooling all year (VESMA guide).

When to use each

Use HDD first. The UK's climate means heating demand dominates for most business estates. The Government's own methodology for non-domestic Operational Ratings corrects main heating energy in proportion to HDD, and it applies a heating-degree-day adjustment with no cooling adjustment (UK Government methodology). If you have a warehouse, workshop or school, HDD against gas or heating electricity will explain most of the seasonal swing.

Add CDD when electricity has a summer bump. If a site has comfort cooling, refrigeration plant or server rooms and electricity use rises in July, test CDD as a second variable. Be cautious, though: CIBSE notes that cooling load does not tend to vary simply or linearly with outdoor temperature, because occupants, equipment and solar radiation also affect it (CIBSE Journal CPD module). A weak CDD relationship is common and does not mean you did something wrong.

Use neither on process loads. A furnace, a compressor line or a recycling shredder responds to production volume, not temperature. Normalize these against output, tonnes or operating hours.

Choosing the base temperature

The base is the outdoor temperature below which a building starts needing heat. Pick it badly and the model fits badly. CIBSE's position is that degree days are valuable because they translate temperature variation into a building-relevant measure, but the base temperature must represent the actual energy balance of the building and its systems (CIBSE TM41).

In practice:

  1. Start with 15.5°C for heating. It is the UK convention and makes your numbers comparable with published tables.
  2. Try 18.5°C for buildings kept warm around the clock, such as care homes or hospitals, and 10°C for well-insulated modern buildings with high internal gains.
  3. Compare the fit (R²) of each base against your own data. The base that gives the cleanest straight line is the better description of your building.
  4. Keep the same base for every period you compare. Changing it mid-series quietly breaks the comparison.

Published UK tables offer heating bases of 18.5°C, 15.5°C and 10.0°C, and cooling bases of 15.5°C, 5.0°C and −20.0°C, which supports consistent comparison between regions and incomplete years (VESMA standard degree days). The Government's Operational Ratings method obtains HDD for the relevant weather region from the building's postcode, so site location matters from the start (UK Government methodology).

The calculation, step by step

The standard approach for heating is a regression with two parameters. Energy use is modeled as base load plus a slope multiplied by degree days. Base load covers energy that does not depend on weather, such as hot water, catering, distribution losses and process loads. The slope reflects the building fabric, heating efficiency and controls (Energy Academy UK).

Step 1: gather matching data

You need consumption per period (monthly bills at minimum, weekly or daily if you have meter data) and degree days for the same periods and the same region. Half-hourly data is better because you can sum it to exact days and strip out out-of-hours periods. It also shows base load clearly at night and weekends.

Step 2: fit the line

Plot consumption against HDD for a baseline year of normal operation. The intercept is your base load; the slope is your weather sensitivity. Test it as ENERGY STAR advises: check how well the model predicts consumption, rather than assuming it does (Benchmarking to Save Energy).

Step 3: compute expected use and compare

Here is an illustrative example with invented numbers, to show the arithmetic. Suppose a site's baseline model is 12,000 kWh per month base load plus 45 kWh per HDD. In a month with 300 HDD, expected use is 12,000 + (45 × 300) = 25,500 kWh. If the meter shows 27,800 kWh, you used 2,300 kWh more than the weather justifies, roughly 9% above expectation. That gap is the thing to investigate: a boiler left on, a controls fault, a changed schedule.

Worked example comparing expected and metered energy use

Step 4: restate to a standard year

To report a weather-corrected annual figure, apply the model to a long-term average HDD total rather than the actual one. UK official statistics do the same in spirit: the greenhouse gas quarterly statistics regress quarterly emissions on quarterly average HDD for each sector and fuel, test statistical significance and goodness of fit, and adjust qualifying series to the 1981–2010 long-term HDD mean (UK greenhouse gas quarterly emissions methodology). Notice the discipline there: only series with a sufficiently strong, significant weather relationship get corrected. Where weather does not explain the variation, no adjustment is imposed. Your own sites deserve the same rule.

Earlier official energy statistics used a different route for some fuels. For coal and petroleum, the former DECC method compared monthly Met Office temperature with the long-term average and applied a fuel- and month-specific factor, producing a series that represents consumption as if every month had average temperature (Temperature correction of energy statistics). It is a useful reminder that there is more than one valid method, as long as it is documented and tested.

Step 4: restate to a standard year

Where weather normalization falls short

Normalization is a time-series tool. ENERGY STAR is explicit that weather-normalized energy adjusts for weather over time but does not by itself adjust for different normal climates between locations, so peer comparisons need climate and operational factors as well (ENERGY STAR). Practical consequences for a multi-site business:

  • Compare each site with its own past, using its own regional degree days, before ranking sites against each other.
  • Normalize for activity too. Opening hours, occupancy, production volume and floor area all move energy use. Weather is one driver among several.
  • Mind the changes. A new extension, a heat pump retrofit or extra EV chargers alter the baseline. If you add vehicle charging, our guide to load management is not the whole story, but the dynamic load balancing checks are worth reading so the new load is planned and then tracked separately in your model rather than blamed on the weather.
  • Do not over-trust the fit. A low R² means the model is not describing your building, whatever the arithmetic says.

What to do with the result

Once you have an expected-versus-actual line, three uses pay back quickly:

  1. Bill checking. An unexplained rise in kWh is a prompt to check the invoice against the meter reads. Our guide on standing charge versus unit rate shows how to separate usage effects from price effects.
  2. KPIs and targets. Weather-adjusted kWh per HDD, or per square meter per HDD, is a fairer indicator than raw kWh. See the energy management KPIs we recommend UK teams track.
  3. Forecasting and budgets. Apply your model to forecast degree days to predict next quarter's use. Our pieces on energy demand forecasting and half-hourly cost forecasts take this further.

For heat-related projects, the same logic proves performance: weather-corrected data is how you judge boiler efficiency or heat pump performance fairly across seasons. Degree days can also normalize heating and cooling emissions against weather, which helps assess building-performance improvements more fairly (SIMAP). If you report carbon, keep weather-corrected and actual emissions clearly labeled, and choose factors carefully; see our note on emissions factor selection.

Our perspective

Most weather normalization fails for a dull reason: the model is built once in a spreadsheet, then nobody updates it. Degree days arrive monthly, bills arrive late and in different formats, and by the time someone compares expected with actual, the month is long gone.

We think the sensible fix is to put the three inputs in one place: validated bills, half-hourly consumption and site location. That way every site has its own baseline, and a gap between expected and actual use shows up while you can still act on it. It is also why we built EnerlyticsAI around bill validation, anomaly detection and half-hourly analytics, with Scope 1, 2 and 3 reporting alongside, and why a free plan covers the core of it. You should not need a budget approval to find out whether a site is wasting energy.

Weather is only one explanation for a change in use. We would rather you ask what the weather justified and then investigate the rest than file every variance under "cold winter".

Chris

Try it on your own sites

If you manage several sites, start by loading their bills and half-hourly data so each one has a clean baseline. EnerlyticsAI has a free forever Starter plan that lets you manage multiple sites without upfront charges. Paid plans are listed on our site: Eco at £49 per month, Growth at £79 per month, Professional at £149 per month and Scale at £299 per month.

Enerlytics AI

You can create a free account, then look at bill validation, half-hourly data and multi-site views to see where use departs from what you would expect.

FAQ

What is the difference between HDD and CDD?

HDD counts degrees below a base temperature and explains heating demand. CDD counts degrees above a base and explains cooling demand. In the UK, HDD usually explains most of a building's weather-related use, with a traditional base of 15.5°C, while CDD is worth testing where you run air conditioning or refrigeration.

Which base temperature should a UK business use?

Start with 15.5°C for heating, because it is the UK convention and matches published tables. Then test 18.5°C for buildings heated constantly and 10°C for well-insulated modern buildings, and keep the one that fits your data best. Use the same base across all periods you compare.

Can I compare two sites once both are weather-normalized?

Not directly. Normalization adjusts one site's use across time, but it does not remove differences in normal climate between locations, or in opening hours and activity. Compare each site against its own baseline first, then compare using intensity measures that also account for floor area or output.

Do I need half-hourly data to normalize?

No. Monthly bills and monthly degree days are enough for a first model. Half-hourly data improves it, because you can separate daytime from night and weekend base load and match consumption to exact days.

Sources