Most of the good opportunities in energy assets are already known to the people running them.

That runs against the usual pitch, where the vendor turns up to point out something the operator missed. It's rarely true. The engineers running a plant know where it's inefficient. They usually have a fair idea what a fix is worth. What they don't have is a number they can defend to a board, or the spare capacity to go and produce one. Their people are rightly focused on running the plant.

That gap, between a hunch and a defensible figure, is where a lot of value sits quietly.

There's a second reason it stays there: everyone assumes proving it properly takes months. It doesn't. Not if the people doing it already know the domain.

Biogas plants are conservative for good reasons. The biology doesn't like surprises, compressors are expensive, and uptime pays the bills. So most operators pick a sensible load and leave it there. Set and forget.

One producer we worked with was running its upgrading compressors at a flat 80%, around the clock, regardless of what power cost. Nothing was broken. The plant ran well. That was rather the problem.

Because electricity prices don't sit still. On a windy, sunny afternoon the day-ahead price can go negative. On a still winter evening it spikes. A plant running flat through both is buying power when it's dear and giving away flexibility it never got paid for. Technically fine, commercially adrift, and nobody could point at exactly where the gap was.

The operator suspected there was money on the table. That suspicion was correct. It just wasn't bankable.

We don't ask anyone to build a control system on faith. So the order of work was deliberate: prove the money exists on paper, then go and capture it.

First a physics-accurate digital twin, built in Wolfram, that behaves like the real plant. An optimiser is only ever as honest as the model underneath it. Then we ran it against real price history and the constraints that actually bind: grid-entry spec, gas quality, compressor duty cycles, and a daily volume commitment that has to be met whatever the market does. Ignore any one of those and you get a number that looks wonderful in a spreadsheet and dies in the field.

That phase was weeks, not quarters, and it was cheap next to the build. We could move at that pace because we weren't learning biogas on the client's time. We already knew what a duty cycle does to an optimisation problem and the ways energy price data misleads you. A generalist software firm would have spent its first month arriving at the point we started from.

The modelling did something a sales deck never could. It put a defensible, plant-specific figure in front of the people who sign off capital, and it won the board's approval to invest. The project stopped being an engineering curiosity and became a budget line.

Only then did we build it. A linear-programming optimiser deciding how hard to run against 15-minute day-ahead prices, on-site solar folded into the same objective, dispatched to the plant over MQTT with the operator able to see and override every decision. The plant now follows the price curve instead of a flat setpoint. The biology is still happy. It's just no longer paying for the privilege of ignoring the market.

The full write-up of that project is here: Turning biogas into bigger profits.

Quantifying first does three useful things.

It tells you whether the opportunity is real at the scale you hoped, which is occasionally a productive disappointment. It surfaces which constraints actually bind, and they're rarely the ones you expected. And it produces the artefact that unlocks the capital: a number derived from your asset, not from a case study about somebody else's.

The objection is always time. In practice modelling first saves it, and it stops you building the wrong thing. The alternative is discovering the margin is thinner than you thought after commissioning, which is an expensive way to learn something a model would have told you in a fortnight.

The hard part of that job wasn't the maths. Linear programming is well-trodden ground. The hard part was earning enough trust in the twin that the operator believed its schedule on a day when the price did something strange, because those are exactly the days the money is made or lost, and exactly the days a fixed-load plant is most tempted to do nothing.

Our clients know their assets far better than we ever will. What we bring is engineers who understand the plant and write the code themselves, so the process thinking and the build are the same job, done by the same people. It means the thing gets chased down properly without pulling your team off running the business.

If you run an asset and you've never seen the cost of standing still written down, that's the first conversation to have. It usually starts with a model, not a sales pitch.

Turning biogas into bigger profits  ·  dispatching an asset against live market signals

Storage for grid support  ·  PMU per-cycle frequency response by dispatching distributed batteries