From one supply to a fleet
Not long ago, a typical industrial site bought power from the grid and kept a generator for outages. Today many energy-intensive plants, from textiles and cement to steel, food and chemicals, run something closer to a small utility: a grid connection, captive gas or diesel engines, perhaps a steam turbine on waste heat, rooftop or ground-mounted solar and, increasingly, battery storage.
Each asset was bought for a good reason. Solar lowers daytime cost. Captive generation protects against outages and high tariffs. Batteries smooth peaks and keep critical loads running. But buying the assets is only half the job. Someone still has to decide, hour by hour, which source should supply the load.
Why manual dispatch leaves money on the table
On most sites that decision is made by shift engineers using fixed schedules, rules of thumb and experience. They are good at keeping the plant running. They are rarely given the tools to minimise cost at the same time, and the problem has become too complex to solve well by hand.
Time-of-use tariffs change the price of grid power by hour and season. Peak windows and demand charges punish short spikes. Gas supply and fuel prices move. Solar output changes with every cloud. Batteries have limits on power, energy and cycles. The common results are familiar:
- Engines running at inefficient part-load because they were started just in case.
- More spinning reserve than the site needs, burning fuel to cover risks a battery could cover.
- Batteries charged and discharged on fixed timers rather than against tariffs and forecasts.
- Solar curtailed or exported at low value while the site imports expensive grid power later in the day.
- Avoidable peaks that set demand charges for a whole billing period.
What a dispatch optimiser does
A dispatch optimiser is software that sits above the existing controls and works out the lowest-cost way to meet the site's load within its technical and contractual limits. It does not replace the power management system, the PLCs or the protection schemes. It tells them, or the operators, what to do next.
In practice it combines four things:
- Forecasts of site load by feeder or process, built from production plans, shift patterns and history.
- Forecasts of solar output from weather data, irradiance and inverter history.
- A cost model for each source: grid tariffs by time band, generator fuel use at different loads and the cost of battery wear.
- Constraints: minimum and maximum loading, ramp rates, reserve requirements, battery state-of-charge limits and grid import or export limits.
Advisory first, automation later
The optimiser re-plans every few minutes as conditions change. Most sites start in advisory mode: operators see a recommended schedule and the reason for it. Once the recommendations have proved reliable, selected decisions, such as battery charging windows, can be automated through the existing controllers, with the plant's engineers keeping control of limits and overrides.
Batteries make the case for this layer sharper. The International Energy Agency calls battery storage “the fastest growing clean energy technology on the market”. A battery's value depends almost entirely on when it charges and discharges: shaving peaks, shifting solar into expensive hours, covering reserve so fewer engines idle on standby, and riding through grid disturbances.
Those uses compete. A battery emptied to shave an afternoon peak cannot cover an evening outage. Getting that trade-off right every day, against changing tariffs and forecasts, is the kind of problem optimisation software handles well and fixed schedules handle poorly.
The data you need
Most of the required data already exists on site, though it is often scattered across systems that do not talk to each other. A year of history is ideal because it captures seasonal patterns, but a pilot can start with a few months and improve as data builds up.
- Interval metering, ideally at 15-minute resolution or better, at the incomer and main feeders. Sub-metering gaps are common and are often the first thing a pilot fixes.
- Generator data: output, fuel flow and running hours, usually available from engine controllers or SCADA over standard protocols such as Modbus or OPC UA.
- Solar inverter data and, where available, on-site irradiance measurements.
- Battery management system data: state of charge, power limits and alarms.
- Tariff schedules, fuel prices and recent energy invoices, so the cost model matches what the site actually pays.
- Production plans and shift patterns, which drive most of the load.
What a typical pilot looks like
A dispatch pilot usually covers one site or complex and runs for 8 to 12 weeks. The first two to three weeks are spent auditing data, closing metering gaps and agreeing the baseline. The optimiser then runs in advisory mode for at least four weeks, while operators compare its recommendations with their own decisions. In the final weeks, operators follow the recommendations within agreed limits and the results are measured against the baseline.
Success measures should be agreed before work starts:
- Energy cost per unit of output, compared with a baseline adjusted for production volume, tariffs and fuel prices.
- The share of recommendations accepted by operators, and the reasons for any overrides.
- Forecast accuracy for load and solar.
- No reduction in supply reliability or power quality.
Savings: an honest range
Savings depend on the site. Typical range: 3 to 8% of energy spend, depending on tariff structure, fuel prices, the mix of assets and how well the site is dispatched today. The basis is the gap between an optimised schedule and existing operating rules, measured against an adjusted baseline.
Sites at the upper end tend to have a wide spread between peak and off-peak prices, several flexible sources and a battery. Sites with a simple fleet, flat tariffs or already disciplined operation will see less. Treat any range, including this one, as a hypothesis to test in the pilot rather than a promise.
There are benefits outside the energy bill too. Fewer engine running hours lower maintenance cost, and a single view of energy use supports cost allocation by line or product and makes energy reporting far easier.
Timing matters
The best moment to design dispatch logic is before a new battery or solar plant is commissioned. Control interfaces, metering points and data access can then be specified once, with the installer, instead of retrofitted. The installer delivers the hardware and its local controls. The optimisation layer that coordinates every source across the site is usually a separate decision, and one worth making early.
Our energy monitoring and dispatch optimisation solution follows this pilot format and draws on our industrial automation and intelligent engineering service. Generator condition data collected along the way can also feed predictive maintenance.
Sources
Published by TashIT on . This article describes general industry practice, not the results of any client engagement. The ranges quoted are indicative and should be tested against your own baseline.
- Energy
- Industrial automation
- Optimisation