Public sector & GeM

Energy baselines for building portfolios

Counterfactual consumption models that make retrofit savings measurable — the arithmetic pay-for-performance programmes stand on.

Layered daily load curves with one flattened after a marked point
Typical pilot
6–10 weeks on one portfolio
Benchmark
ASHRAE Great Energy Predictor III
Benchmark host
ASHRAE
Sector
Public sector & GeM
What this is — and is not

This page describes a problem class and the architecture we deploy for it, written against a named public benchmark — the open dataset or competition where the world's data scientists have tested approaches to this exact problem against a hard metric. It is not a client engagement, and the benchmark's results, prizes and rankings belong to its host and participants, not to us. Client work is confidential and is only ever published with written permission.

The problem

What it costs when this goes unsolved.

An energy-efficiency retrofit pays for itself only if you can prove what consumption would have been without it. That counterfactual baseline is a modelling problem, and it carries money: measurement-and-verification is what lets an ESCO contract on performance, a portfolio owner rank buildings by savings potential, and a public programme audit what it paid for.

The data reality

Metered consumption across a building portfolio is long, wide and dirty — electricity, chilled water, steam and hot water meters at hourly grain, with gaps, unit errors and whole periods of mislogged data. The reference benchmark for this problem spans 20 million training points across 1,448 buildings; the published lesson from it is that disciplined data cleaning and per-meter decomposition beat clever architectures.

How we build it

The architecture, stage by stage.

01

Meter-level decomposition

Separate models per meter type and site cluster, because steam does not behave like electricity and a campus does not behave like an office block.

02

Weather-normalised gradient boosting

LightGBM ensembles over weather, calendar and occupancy features — the approach the benchmark's peer-reviewed analysis found strongest at portfolio scale.

03

Anomaly handling before modelling

Systematic detection of meter faults, unit errors and dead periods, documented so the baseline survives an auditor's challenge.

04

Savings attribution with intervals

Retrofit savings reported against the baseline with uncertainty bounds — the honest number a performance contract can settle on.

Model families on this problem class: LightGBM ensembles · Weather normalisation · Time-series anomaly detection · Portfolio dashboards.

What you get

What an engagement hands over.

Everything below goes in the scope document before you sign it, with a fixed price or a rate with a ceiling — the same terms as every other engagement in the catalogue.

  • Baseline models per building and meter, weather-normalised
  • A measurement-and-verification report format aligned with IPMVP practice
  • Portfolio ranking by expected savings, for capital planning
  • Anomaly flagging that pays for itself in found meter faults alone

Provenance

The benchmark behind this page.

ASHRAE Great Energy Predictor III, run by ASHRAE, is the public proving ground for this problem class. The figures below are the host's, cited as context for how seriously this problem is tested in the open — they are not our results and we do not claim them.

Benchmark scale
3,614 teams from 94 countries · US $25,000 prize
Data scale
20M+ training points across 1,448 buildings at 16 sites — the largest building-energy ML benchmark ever run
Metric
RMSLE, with peer-reviewed outcome analysis
Reference
Benchmark page ↗

Next step

Thirty minutes on whether this fits your problem.

Bring the constraint — the regulator, the data boundary, the latency budget. If your data cannot support this build yet, the call will conclude with what to fix first, not with a proposal.