Industrial Demand Response: Peak Shaving and Tariff Optimisation

Industrial demand response means adapting when and how much a plant consumes in order to reduce its electricity bill and its exposure to market prices, without sacrificing production. Throughout this article we look at how tariff periods and contracted capacity work, what peak shaving is, how to assess the real flexibility of your processes and what must be measured before moving a single load.

Managing demand, not just generating

For decades, the only way a manufacturer could act on its electricity cost was to negotiate the kWh price or install its own generation. Demand was taken as fixed: the plant consumes what it consumes and the power system takes care of supplying it. That approach no longer holds. In a system with ever more renewable generation, the price of electricity varies widely over the day, and the structure of network charges penalises consumption during the hours when the system is under most stress. The practical consequence is that two plants with the same annual consumption can pay very different bills depending on the hourly profile of that consumption.

Managing demand means treating the consumption profile as one more operational variable, on a par with production scheduling or maintenance. It does not mean consuming less in total (that is energy efficiency, which has its own path), but consuming at a different time or with a different shape. The two disciplines complement each other: efficiency reduces the kWh, and demand management reduces what each of them costs. A plant that already works on efficiency with an ISO 50001 style approach is halfway there, because it has data and a culture of energy monitoring.

In the broad sense, demand response ranges from internal decisions (shifting a process to off-peak hours) to participation in flexibility services that pay consumers for reducing load when the system operator requests it. This guide focuses on the former, which is where almost any plant can start tomorrow: peak shaving and tariff optimisation with the means it already has.

Tariff periods and contracted capacity, at a conceptual level

To understand what can be optimised, it helps to separate the three big blocks that make up the electricity cost of a grid-connected industrial site:

In Spain, industrial network access tariffs divide the day into several tariff periods (six periods, P1 to P6, in the higher-voltage tariffs), with network charge prices that can differ by an order of magnitude between the most expensive period and the cheapest. The period calendar varies by season and zone, so any serious analysis starts from the specific calendar that applies to the supply point. The underlying idea, however, is stable: the middle hours of winter and summer working days are expensive, while nights, weekends and much of August are cheap.

Contracted capacity works differently. The plant contracts a maximum number of kW per period and pays for that maximum every month, whether it uses it or not. If consumption exceeds the contracted level, the maximum-demand meter records it and the retailer bills excess charges, which in industrial tariffs are calculated from the quarter-hourly readings and can be expensive if peaks are frequent. This creates a classic optimisation problem: contracting too much means paying every month for idle capacity; contracting too little means paying penalties. The solution is to know the real demand curve at quarter-hourly resolution and decide with data, not with safety margins inherited from ten years ago.

What peak shaving is

Peak shaving means reducing the maximum power the plant draws from the grid, normally by shifting or modulating loads so that they do not coincide with each other or with the expensive periods. A plant's demand peak is almost never caused by a single machine: it is caused by coincidence. A furnace starting up while the compressors are at full load and the HVAC equipment is responding to the midday heat produces a peak that none of the three would generate on its own.

That is why peak shaving is, above all, a problem of temporal coordination. The usual techniques, from least to most complex:

1. Staggering start-ups. Equipment with high inrush or start-up peaks (large motors, furnaces, cold rooms after a stoppage) is started in sequence, not all at once. It is the cheapest measure and often the most effective.
2. Shifting schedulable loads. Processes that do not require a specific time slot (pumping to a tank, producing refrigeration or compressed air with storage, charging batteries of internal vehicles, certain batch treatments) are moved into off-peak periods.
3. Modulating interruptible loads. Equipment that tolerates brief reductions without affecting the process (building HVAC, ventilation, industrial refrigeration with thermal inertia) is turned down during peak minutes.
4. Load shedding with priorities. A control system watches the quarter-hourly power and, when the forecast points to an excess, disconnects loads according to a priority list agreed with production.

The economic result of peak shaving arrives via two routes: it allows less capacity to be contracted in the expensive periods, and it eliminates excess charges. The energy-term route (moving kWh from P1 to P6, for example) is conceptually different but rests on the same capabilities: knowing which loads are flexible and being able to act on them.

Flexibility of industrial processes

The decisive question is not how much power each piece of equipment consumes, but how much temporal freedom it has. A flexibility map classifies the plant's loads into four categories:

The key to modulable loads is implicit storage. A well-insulated cold room is, in electrical terms, a thermal battery: it can be cooled one degree extra during off-peak hours and allowed to drift during the peak without leaving the product's admissible range. The same goes for a water tank, a compressed air receiver or an intermediate silo between two process stages. Identifying these hidden stores usually uncovers more flexibility than production believes it has, because nobody had looked at them through that lens.

The limits of flexibility are set by quality and safety, which is why the map cannot be drawn by the energy department alone. Every schedulable or interruptible load needs sign-off from whoever answers for the process, with explicit conditions: admissible temperature range, maximum duration of the interruption, loads that are never touched. That written agreement is what later allows automation without nasty surprises.

The four levers of action

With the flexibility map done, the measures line up in four levers, from least to most investment:

1. Contract optimisation. Reviewing contracted capacity by period against the real quarter-hourly curve of the last year. It is common to find capacity levels inherited from a plant configuration that no longer exists. Implementation cost is close to zero; the saving depends on the mismatch found.

2. Operational rescheduling. Changing setpoints and schedules: staggering start-ups, moving schedulable batches into off-peak, producing refrigeration in advance. It requires no equipment investment, but it does require operational discipline and, to hold over time, some degree of automation.

3. Automatic demand control. A system that measures in real time, projects the power of the current quarter-hour and acts on the modulable and interruptible loads according to priorities. It is the lever that turns peak shaving into something reliable instead of depending on someone watching a screen.

4. Flexibility assets. Batteries, expanded thermal storage, on-site generation. This is the highest-investment lever, and it only makes sense to size it once the previous three have shaved whatever could be shaved without spending, because every peak eliminated through operations is battery capacity you do not have to buy.

What needs measuring to get started

None of the levers above works without data. The minimum starting point is the load curve at the grid connection point at quarter-hourly resolution, which the fiscal meter already records and which the distribution company or the retailer can provide. With just that curve and the period calendar you can already audit the contracted capacity and quantify excess charges. But to act on specific loads you need to go one level down: submetering on the relevant lines or machines, because the overall curve tells you when there is a peak but not who causes it.

A reasonable starting installation includes:

• The quarter-hourly curve at the connection point, with at least twelve months of history to capture seasonality.
• Power analysers on the switchboards of the large loads, or those suspected of causing peaks: compressors, refrigeration, furnaces, HVAC.
• Process context: machine states, shifts, production orders. Without this context, a peak is a mystery; with it, it is an explainable and avoidable event.
• A platform that brings all of the above together on a single timeline and lets you answer questions like "what was running on Tuesday at 12:15".

This is exactly the territory of connected energy management: continuous monitoring with submetering, integrated with the plant's operational data. Without that foundation, demand response remains an exercise in bills; with it, it becomes a measurable operating routine.

Worked example: shaving a compressor peak

A simplified case with round numbers to illustrate the mechanism. A plant has a base demand of 800 kW on the morning shift. Its compressor room adds up to 240 kW and starts in cascade at 6:00, just as the main line also starts. The maximum-demand meter records peaks of 1,150 kW, so the plant keeps 1,200 kW contracted in all peak periods to avoid excess charges.

Analysis of the quarter-hourly curve with submetering reveals two things. First: the 1,150 kW peak lasts less than twenty minutes and only occurs at shift start-up, through the coincidence of compressors and line. Second: the receiver and the air network have enough capacity to sustain air demand for a few minutes with the compressors modulated.

The measures, in order:

1. Bring the compressor start-up forward to 5:30, in a cheaper period, so that the air network is pressurised before the line starts.
2. Stagger the compressor cascade with five-minute delays between stages.
3. Configure the control so that, if the projected power of the quarter-hour exceeds a threshold of 1,000 kW, it temporarily modulates the back-up compressor.

With the peak consistently contained below 1,000 kW for several months, the plant can consider reducing contracted capacity in the expensive periods by around 200 kW. The exact saving depends on the tariff and the prices in force, and there is no point stating it here in euros; what transfers from the example is the method: submetering to locate the cause of the peak, an implicit store (the compressed air) as a buffer, temporal coordination as the main measure, and an automated threshold as the safety net. None of the three measures touched the production process.

Relation to self-consumption and storage

Demand response and photovoltaic self-consumption reinforce each other, but it is worth understanding where they help each other and where they do not. An industrial self-consumption installation reduces the energy bought from the grid during sunlight hours, which largely coincide with expensive periods: excellent for the energy term. It is not, however, a guarantee against demand peaks, because the peak can occur at a moment of low solar output (a 6:00 start-up, a cloudy day). Sizing contracted capacity counting on the photovoltaics as if it were firm capacity is a classic error.

This is where demand management multiplies the value of self-consumption: shifting flexible loads towards solar production hours increases direct self-consumption, which is the highest-value energy of the installation, and at the same time reduces purchases in expensive periods. A plant that already has its flexibility map done can raise its self-consumption rate by several points without adding a single panel.

Battery storage closes the triangle. A battery can do pure peak shaving (discharging during the peaks), price arbitrage (charging off-peak, discharging at peak) and mop up solar surpluses that would otherwise be exported. Its sizing depends on the shape of the peaks: short, sharp peaks are solved with relatively little capacity; a high, sustained demand lasting hours requires large batteries, and operational measures are usually the more reasonable alternative. That is why the order matters: first measure, then squeeze the operational flexibility, and only then size generation and storage on the already-optimised curve.

Common mistakes

• Optimising on the bill instead of on the curve. The bill aggregates; the quarter-hourly curve explains. Capacity decisions taken from bills alone tend to leave money on the table or generate excess charges.
• Cutting contracted capacity before controlling the peaks. The fixed saving on the capacity term can be wiped out by excess charges if the plant has no mechanisms to contain demand.
• Ignoring production. A load-shedding plan that production has not signed off gets deactivated in the first week. The priority list has to be born as an agreement.
• Treating it as a one-off project. The plant changes: new equipment, different shifts, another product mix. Without continuous monitoring, today's optimisation is the mismatch of two years from now.
• Starting with the battery. Flexibility assets are sized on the curve already worked through operations, not on the raw curve.

Frequently asked questions on industrial demand response

At CAPTIA Energy we help industrial plants turn their consumption profile into a savings lever: quarter-hourly metering and submetering, a flexibility map, peak control and integration with self-consumption. If you want to know how much flexibility your plant is hiding, talk to our team.