Automation firms: models, optimisers and simulation, few experts

Protection and optimisation of industrial grids and microgrids

What happens. Customers want their site network of generators, storage, loads and grid connection to be controlled and protected as one system. The supplier has to show in simulation that the design holds before anything is built.

What it costs. Projects need specialists in both power systems and optimisation, and these are rare. Without them the supplier either offers simple rule-based control or takes on risk it cannot judge.

Why the usual fix fails. A rule-based controller is easy to deliver, but it ignores prices, forecasts and the limits of the connection. Hiring a full-time expert for a few projects a year is rarely possible.

How a fixed-price remote diagnosis checks it. From single-line diagrams, load profiles and component data I check which control and optimisation approach fits the site, and what the simulation must show to be credible. The fixed-price remote diagnosis gives a project team an outside view of design and simulation before it commits.

Background: Economic predictive control

Plant models must be built from documents and historical data

What happens. For an optimiser to work, someone has to build a model of the energy and process systems of a customer’s site. The only inputs are piping diagrams, data sheets, operating manuals and years of logged data of mixed quality.

What it costs. Model building eats engineering time, and a model fitted badly gives an optimiser that suggests things the plant cannot do. Projects slip or are delivered with simplified models.

Why the usual fix fails. Generic models from a tool library do not capture the particular plant. Fitting a purely statistical model to history ignores the physics and fails outside the range it has seen.

How a fixed-price remote diagnosis checks it. I use physical models fitted to the available data and check where the data is too thin to support them. The fixed-price remote diagnosis shows how far plant documents and logs go toward a usable model, and what missing measurements would change that.

Background: Replacing hand-tuned parameter tables by a model

Turbines, boilers, cogeneration and steam must become one optimisation

What happens. A site has turbines, boilers, combined heat and power units, steam networks and storage, all with their own limits, efficiencies and costs. The task is to schedule them together against prices and demand.

What it costs. Without an overall plan, each unit is run by its own rules and the site pays for steam that is let down, heat that is wasted and units that are started and stopped too often.

Why the usual fix fails. A spreadsheet or a simple rule set for the control room cannot hold start-up limits, minimum loads, storage and price forecasts at once. Software off the shelf needs a good model of each unit, which is the hard part.

How a fixed-price remote diagnosis checks it. I take unit data, performance curves and operating logs and set up the optimisation problem with its limits and costs. The fixed-price remote diagnosis tells you whether the model of your site is good enough to optimise on and what the plan would have saved on past days.

Background: Economic predictive control

Process simulators and virtual commissioning around the control system

What happens. Customers expect the control programme to be tested against a simulated plant before the real one is built. That takes a dynamic model, a link to the control system, and someone who can judge whether the simulation behaves like the plant.

What it costs. Without that, faults are found during start-up on site, where every day is expensive and every change is risky. With a poor simulation, the team trusts a result that does not hold.

Why the usual fix fails. Simple simulations that only play back signals prove that the logic runs, not that the control works. A full simulator of every loop is too much work for most projects.

How a fixed-price remote diagnosis checks it. I look at design data, equipment data and the control concept and decide which parts of the plant need a dynamic model for the test to mean something. The fixed-price remote diagnosis shows what a useful simulation would include for your project and which start-up risks it can really remove.

Advanced process control and optimisation for pulp and paper mills

What happens. Automation suppliers are asked to offer advanced process control for pulp and paper mills: predictive control of the stock preparation, the paper machine and the drying section, with computed qualities where the lab is too slow.

What it costs. The work needs people who know both the paper process and predictive control. Without them the supplier cannot take on the work, or the controllers are commissioned and then switched off by operators.

Why the usual fix fails. A conventional control loop for each variable fights the interactions in a paper machine. Recruiting a specialist team takes years.

How a fixed-price remote diagnosis checks it. From mill logs (basis weight, moisture, flows, consistency, lab results) I check how well qualities can be computed and where predictive control would act. The fixed-price remote diagnosis gives the supplier a first view of what a project in a mill would involve.

Background: Predictive control of process plants