Defining the technologies: IPC, CPC and the art of the query.

“Clean energy” is not a definition; it is a mood. Ask three fund managers what belongs in it and you will get three different portfolios and at least one heated argument. Our starting point is different: if the methodology reads patents, the theme must be defined in the language of patents. That language exists, it is public, and it is maintained by the patent offices themselves.

The world’s filing cabinet for inventions

Every patent application on earth is classified by an examiner into the International Patent Classification (IPC), maintained by the World Intellectual Property Organization, and most also carry codes from the finer-grained Cooperative Patent Classification (CPC) developed by the European and US patent offices. Think of these as the postcode system of invention: hierarchical, globally consistent, and assigned by an independent examiner rather than by the company doing the inventing.

The precision is remarkable. Within the broad class for gas separation, code B01D53/14 designates absorption processes — the chemistry behind capturing CO₂ from flue gas. The CPC even carries a dedicated branch, Y02, created specifically to tag climate-change mitigation technologies. A company cannot lobby its way into these codes; an examiner assigns them based on what the invention actually does.

From codes to a taxonomy

On this scaffolding we build our own taxonomy. The decarbonisation universe is organised in three layers — theme class, subclass, technology — covering carbon capture and storage, efficiency gains (buildings, motors, energy distribution and storage, low-carbon transport), and energy substitution (renewable generation, transport electrification, sustainable aviation and fuels). At the working level this resolves into roughly seventy narrowly defined technologies, each with a written definition of what it covers and, just as important, what it does not.

The narrowness is the point. “Energy storage” is too broad to identify a leader in anything; “grid-scale battery storage” and “hydrogen storage” are separate technologies with separate leaders, and the taxonomy keeps them separate.

Why keywords alone fail: the wave power problem

Take wave power — machines that harvest mechanical energy from ocean waves. Search the patent record for “wave energy” and you will drown in the wrong waves: radio waves, radar, ultrasound, optical waveguides, seismic imaging. Half of modern physics travels in waves, and every wireless router on earth handles “wave energy” of a sort. Keywords cannot tell an oscillating water column from a Wi-Fi antenna.

Classification codes can. The CPC places machines driven by wave or tidal energy in their own branch of F03B — liquid machinery — a code no wireless invention will ever carry, with the climate-mitigation tag Y02E10/32 alongside. Anchor the query in those codes and the electromagnetic noise never enters; the keywords are then free to do their proper job of separating tidal turbines and wave energy converters from the rest of marine engineering.

Anatomy of a query

Each technology becomes a search query against the global patent record in Patsnap. A query has four working parts, and each earns its place:

Classification codes form the backbone — the objective, examiner-assigned definition of the technical field.
Keyword inclusions sharpen the scope where codes alone are too broad, using the vocabulary of the technology itself.
Exclusions — both keywords and classification codes — strip out false positives: the neighbouring fields, legacy applications and noise that share codes or vocabulary with the target technology but do not belong to it.
Filters keep only what matters for security selection: patents that are legally active or pending, held by companies rather than universities or individuals, within a defined time window.

A simplified extract from the actual tidal and wave power query shows the shape:

IPC_CPC:(F03B13/10 OR F03B13/12 OR … OR Y02E10/32)
AND ("tidal turbine" OR "wave energy converter" OR "oscillating water column" OR "point absorber" …)
NOT ("offshore wind" OR "ship propulsion" OR radar OR wireless OR ultrasound OR "seismic wave" …)
AND legal status: active or pending, assignee type: company

The exclusions are where most of the craft sits — and they work in both directions. This query strips out the neighbours that share its waters (offshore wind, ship propulsion) as well as the false waves (radar, ultrasound, seismic). Meanwhile the hydropower query, one branch over, explicitly excludes “wave power” and “ocean current” — because a dam turbine and a wave energy converter share machinery codes but have different leaders, and the taxonomy refuses to blur them. Each query is, in effect, a formal statement of what we believe the technology to be — written down, testable, and applied identically to every company in the world.

Keeping it honest

Classification systems are living things — the patent offices release updates, new codes appear as technologies mature, and the two systems (IPC and CPC) do not always map neatly onto each other. The taxonomy is therefore maintained, not set in stone: each release is reprocessed, IPC codes are matched to their closest CPC equivalents using semantic similarity models, and every mapping below a confidence threshold is flagged for manual review before any query is rebuilt. An audit pass checks the taxonomy for gaps and low-confidence matches before it touches the live process.

None of this is glamorous. But it is precisely this unglamorous layer that makes everything downstream defensible: when a company scores well in grid-scale storage, we can show exactly which patents drove it and exactly why those patents belong to that technology. Definitions first; scores later.