Energy Monitoring System Cost & Rollout for Thai Factories

Electricity prices in Thailand have been trending upward through 2026, and for the September–December period the Energy Regulatory Commission has put a proposal out for public hearing that would raise the Ft (fuel adjustment charge) from 0.1623 THB/kWh to as much as 0.9482 THB/kWh (undecided at the time of writing). If the highest option is adopted, a plant consuming 250,000 kWh a month faces more than 2 million THB in additional cost per year. You cannot set the tariff. What you can change is when, where, and how much you consume — and an energy monitoring system is what makes that possible. This article covers the structure of Thailand’s industrial tariff, a full bill breakdown for a model plant, a savings estimate, a five-layer cost breakdown, and a 90-day roadmap to get a system live.

What is actually happening to factory electricity costs in Thailand in 2026

Recent tariff movement and the proposed Ft increase (not yet confirmed)

Start with the facts. The average electricity tariff published by Thailand’s Energy Regulatory Commission (ERC) was 3.88 THB/kWh for the January–April 2026 period and 3.95 THB/kWh for May–August, an increase of 1.8% over the previous period. The May–August figure breaks down into a base tariff of 3.78 THB/kWh plus an Ft of 0.1623 THB/kWh (16.23 satang).

The next period is where it gets uncomfortable. At public hearings held from 13 to 20 July 2026, four options were tabled for the September–December Ft, ranging from 0.1623 to 0.9482 THB/kWh. The corresponding all-in tariff options are 3.95 / 4.20 / 4.25 / 4.73 THB/kWh. At the time of writing, which option will be adopted has not been decided.

It is worth checking how those numbers relate to each other (the following arithmetic is ours):

• Hold-steady option: 3.78 + 0.1623 = 3.9423 THB/kWh → corresponds to the published 3.95 THB/kWh
• Maximum option: 3.78 + 0.9482 = 4.7282 THB/kWh → corresponds to the published 4.73 THB/kWh
• Ft multiple: 0.9482 ÷ 0.1623 = roughly 5.8x

One caveat on precision: simply adding the base tariff and the Ft gives 3.9423 THB/kWh, which does not match the ERC’s published average of 3.95 THB/kWh at the second decimal place. The difference is attributed to the allocation of service charges and to rounding. For the estimates in this article we build the bill up from its components — base tariff, TOU energy charges, demand charge, and Ft — so that every line item can be traced.

“The rate went up” is not an action plan

The typical reaction inside a plant when a tariff increase hits the news is a company-wide announcement asking everyone to save electricity, followed by someone switching off half the lights and turning up the office air conditioning. But the overwhelming majority of a factory’s electricity goes into production equipment. Turning off office lights barely moves the invoice.

To decide what to actually do, you have to break the invoice down into the factors that get multiplied together to produce the number at the bottom. Thailand’s industrial tariff has four main components:

1. Base tariff
2. Ft (fuel adjustment charge)
3. TOU energy charge (a per-kWh rate that differs between on-peak and off-peak)
4. Demand charge (a monthly charge per kW)

Of these four, the base tariff and the Ft are entirely outside a plant’s control — you cannot move either by a single satang. What a plant *can* change comes down to three things: *when* you consume (TOU), *how high* you let your peak go (demand), and *how much* you consume in total. An energy monitoring system — variously called a power monitoring system or an EMS (energy management system) — earns its keep on exactly those three levers.

Breaking the industrial tariff into four parts

1. Base tariff — the cost of generation, transmission, and distribution

The base tariff is 3.78 THB/kWh for the May–August period and covers the underlying cost of generation and network assets. It is set through policy review and is not something plant operations can affect.

2. Ft (fuel adjustment charge) — fuel prices and exchange rate movement

The Ft is the adjustment mechanism that passes fuel prices, exchange rates, and changes in the generation fuel mix through to the tariff. It is reviewed every four months and stood at 0.1623 THB/kWh for May–August. Because it applies directly to every kWh consumed, the more electricity you use, the harder it hits. We quantify the impact of the maximum September–December option later in this article.

3. TOU energy charge — the same kWh costs 1.6x more depending on the hour

TOU (Time of Use) is time-of-day pricing. In Thailand the periods are defined as follows.

For industrial Type 3 service at a supply voltage of 12–24 kV, the energy charges (excluding Ft) are:

The gap is 4.1839 − 2.6037 = 1.5802 THB/kWh, a ratio of 4.1839 ÷ 2.6037 = approximately 1.61x (our calculation). Put plainly: the identical kilowatt-hour costs 1.6 times more at 10 a.m. on a Monday than at 10 a.m. on a Saturday. That gap is the entire basis for the load-shifting measure discussed later.

4. Demand charge — billed on kW, not kWh

The demand charge is not based on consumption (kWh). It is based on the highest instantaneous rate of use (kW) you reach during the month. The rate varies with supply voltage.

For the same plant with a maximum demand of 800 kW, that is 800 × 210.00 = 168,000 THB/month below 12 kV, 800 × 132.93 = 106,344 THB/month at 22–33 kV, and 800 × 74.14 = 59,312 THB/month at 69 kV and above (our calculation). The spread between the below-12 kV rate and the 12–24 kV rate is 210.00 ÷ 132.93 = approximately 1.58x. In older plants that have grown through successive expansions, the supply voltage and tariff class sometimes no longer match the actual load profile — and a contract review alone can move the number.

For reference, industrial Type 3 covers contracted demand of 30–999 kW, and Type 4 covers 1,000 kW and above (TOU mandatory).

The two rules that decide your demand charge: the 15-minute interval and the 70% ratchet

This is the section we most want readers to take away. In practice, demand monitoring and demand control succeed or fail on whether the team understands these two rules.

One 15-minute average, and your whole month is decided

Thailand’s demand charge is set by a single data point: the highest 15-minute average power reading in the month. It is not the instantaneous peak — it is a 15-minute average. But turn that around and the implication is uncomfortable: if a few large loads happen to overlap for 15 minutes, that one interval becomes your billed demand for the month.

It helps to count how many intervals there are (our calculation):

• Per day: 24 hours × 60 minutes ÷ 15 minutes = 96 intervals
• Per 30 days: 96 × 30 = 2,880 intervals

So one interval out of 2,880 determines 106,344 THB/month for the model plant described below. No matter how well you run the other 2,879, a single overlap wipes the slate.

The circumstances that produce that one interval are remarkably predictable:

• Monday morning start-up: chillers, HVAC, compressors, and furnaces that were shut down over the weekend all start at once
• Return from the lunch break: idled equipment comes back simultaneously
• A furnace or dryer ramping up to temperature at the same moment moulding machines start
• A very hot afternoon: air conditioning and chillers both run flat out
• Trial runs, commissioning, and unplanned extra production

None of these are anyone doing anything wrong. It is purely a matter of timing coinciding. Which is exactly why real-time demand monitoring, paired with a written rule about what gets shed when the alarm sounds, can cut the charge without a single baht of new equipment.

The 70% ratchet — the last 12 months follow you into this month

The second rule is the ratchet. Billed demand is determined as follows:

“`javascript

Billed demand = MAX(actual demand this month, highest demand in the last 12 months × 70%)

“`

This is the rule readers most often miss. Once you set a high peak, that peak establishes a floor at 70% of its value for the next 12 months, no matter how well you perform.

Take the model plant introduced below, whose highest demand over the last 12 months was 800 kW (our calculation):

• Ratchet floor: 800 × 70% = 560 kW
• If actual demand this month is cut to 700 kW: MAX(700, 560) = 700 kW → the full reduction flows through
• If actual demand this month is cut to 500 kW: MAX(500, 560) = 560 kW → you cut 300 kW in reality, but only 800 − 560 = 240 kW shows up on the bill

In money terms:

In Case B you cut 300 kW of real demand but get paid for 240 kW. The missing 60 kW — worth 300 × 132.93 − 31,903 = 7,976 THB/month (our calculation) — is unrecoverable while the ratchet holds. It is not lost forever: once 12 months roll past and the trailing maximum resets to a lower value, the floor comes down with it. The point is that the benefit arrives with a lag, and you need to know that before you take the business case to the board.

Three consequences of the ratchet for your investment case

1. Not creating a big peak is the highest-value action available. One 800 kW event keeps a 560 kW floor in place for the next 12 months. Simply having demand in mind before a commissioning run or an unscheduled shift changes twelve months of fixed cost.
2. Measure results using billed kW, not actual kW. The peak on your dashboard can fall while the invoice does not. Make sure “billed demand (after ratchet)” is a KPI on the dashboard, not an afterthought.
3. Build the ratchet floor into the payback calculation. If you total up projected kW savings on an actual-demand basis, you will overstate year-one benefit. Check the trailing 12-month maximum *first*, then run the numbers.

You can do this check today, before any system exists. Pull the last 12 invoices, find the highest demand, and multiply by 0.7. That number is your plant’s floor — the level below which your bill will not go.

A model plant: where the money actually goes

Abstractions do not support decisions, so let us take a concrete model plant and take its invoice apart.

Model plant parameters

A load factor of 43.4% means that, against the theoretical consumption you would reach by holding 800 kW continuously for 30 days, the plant actually uses 43.4% of it. That is a typical level for a day-shift operation with some equipment running overnight and at weekends — neither unusually low nor unusually high. As you will see shortly, this load factor is what drives the demand charge’s share of the bill.

Monthly bill breakdown (Ft = 0.1623 THB/kWh; service charge and VAT excluded)

An effective rate of 4.14 THB/kWh sits near the bottom of the 4.10–5.50 THB/kWh range cited for Thai factories in the May–August period (TOU and demand charge included). That is because the load factor is 43.4% — there are enough kilowatt-hours to spread the demand charge across. Plants with a lower load factor (meaning big loads for short periods) divide the same demand charge by fewer kWh, and their effective rate lands higher.

The full composition, for completeness (our calculation):

Why this contradicts the “demand charge is 20–30% of the bill” rule of thumb

The commonly quoted figures are that demand charges make up 20–30% of a factory’s bill and energy charges 50–60%. In our model plant the demand charge is only about 10.3%.

That is not a contradiction — it is a difference in load factor. The demand charge is maximum kW × rate, and it is incurred regardless of how many kilowatt-hours you consume. Therefore:

• High load factor (equipment running steadily over long hours): the kWh denominator is large, so the demand charge’s share falls
• Low load factor (big loads for short bursts): the kWh denominator is small, so the demand charge’s share rises

For the same reason, energy charges come out higher than the 50–60% rule of thumb suggests (85.8% in the model plant): the larger the kWh denominator, the less weight the kW-based charge carries. The 20–30% rule of thumb describes low-load-factor plants. Which one you are takes about five minutes to establish: pull the kWh and the kW off your invoice and compute your load factor. The lower it comes out, the higher demand peak-shaving should sit on your priority list. Getting this first-pass diagnosis wrong — and jumping straight to “let’s do solar” or “let’s do VFDs” — is how ROI becomes unpredictable.

What happens if the Ft goes to the top of the range (unconfirmed scenario)

Suppose the September–December Ft lands at the maximum proposal of 0.9482 THB/kWh. Here is what that does to the model plant. To repeat: whether this option is adopted is undecided at the time of writing.

Roughly 196,000 THB a month, or 2.36 million THB a year. No capital investment, no additional headcount — just a fixed cost that lands on the P&L.

One nuance: on published-tariff terms the increase is (4.73 − 3.95) ÷ 3.95 = approximately 19.7% (our calculation), yet the model plant’s increase comes out slightly lower at about 19.0%. That is because its effective rate of 4.14 THB/kWh sits above the published average — a high on-peak share plus a demand charge on top — so the denominator is larger. The Ft applies only to kilowatt-hours, never to the demand charge portion (about 10.3% of the bill). A plant with a low load factor and a high demand-charge share therefore sees a *proportionally* smaller Ft impact. But the absolute increase is the same 196,475 THB/month for anyone consuming 250,000 kWh. Judge this on the baht, not the percentage.

Three levers an energy monitoring system unlocks — and what they are worth

Installing an energy monitoring system does not, by itself, reduce your electricity bill. The accurate framing is that it lets you execute three specific measures on the basis of facts instead of opinions. Here they are in order.

Lever 1: Demand peak shaving (800 kW → 700 kW)

This is the case where maximum 15-minute demand is brought down from 800 kW to 700 kW, a reduction of 12.5% (100 ÷ 800).

Removing 100 kW does not mean shutting equipment down. It means stopping things from happening at the same moment. Specifically:

• Stagger start-up sequences: on Monday morning, bring compressors, chillers, HVAC, and furnaces online 10–15 minutes apart instead of together
• Move furnace and dryer ramp-up so it finishes before the on-peak window opens
• Set predictive alarms: alert while the 15-minute interval is still running, at the point where the current trajectory will exceed the target kW (an alarm that fires after the interval closes is useless)
• Decide the shed order in advance: write down, on paper, which equipment gets curtailed first, ranked by the least impact on quality and safety

As noted above, pushing all the way down to 500 kW runs into the 560 kW ratchet floor and yields only 31,903 THB/month (31,903 × 12 = approximately 382,836 THB/year). For this estimate we use the realistic and achievable 700 kW target.

Lever 2: Shifting load from on-peak to off-peak (20,000 kWh/month)

Here, 20,000 kWh of the 150,000 kWh consumed on-peak — 20,000 ÷ 150,000 = approximately 13.3% of on-peak consumption (our calculation) — is moved into the off-peak window.

Same work, same output, different clock — worth roughly 379,000 THB a year. The loads that realistically shift are:

• Thermal storage on chillers and refrigeration plant (make cold at night, use it during the day)
• Air receiver charging on compressors; pumping to ground and elevated water tanks
• Batch thermal processes: heat treatment, annealing, drying
• Cleaning and surface-treatment batches; mould preheating
• Moving selected processes to a night shift (subject to a proper comparison against labour cost and overtime regulations)

The critical constraint here is that you cannot choose what to shift unless you know which equipment consumes how many kWh during the on-peak window. If your only measurement point is the incoming supply, this conversation cannot even begin. That is the concrete reason equipment-level power measurement is required.

Lever 3: Continuous energy management (ISO 50001-style, 4% per year)

Plants running a functioning EnMS (energy management system) have been reported to sustain roughly 4% annual energy reduction for more than ten consecutive years (as reported by the US DOE Better Plants programme). Applying that level to the model plant:

The content of that 4% is unglamorous and cumulative: finding and repairing compressed-air leaks, eliminating standby loads, right-sizing compressor discharge pressure, revisiting chiller setpoints and cooling water temperature, enforcing shutdown rules for non-production hours, and prioritising small capital items such as VFD retrofits and LED conversion. Every one of them depends on being able to see, numerically, *where* the waste is before you can rank what to fix.

Air leaks and equipment degradation often show up first as changes in the electrical waveform. Once you hold consumption trends at equipment level, the same data serves not only energy reduction but early detection of equipment faults. Evaluating this alongside a predictive maintenance system materially improves the return on the same measurement infrastructure.

The three levers combined, and payback

Expressed as a percentage of the annual electricity bill (our calculation):

• Annual electricity cost: 1,034,874 × 12 = 12,418,488 THB/year
• Reduction: 1,035,564 ÷ 12,418,488 = approximately 8.3%

And, tying back to the Ft scenario at the top of the article:

• 1,035,564 ÷ 2,357,700 = approximately 43.9%

In other words, of the 2,357,700 THB annual cost increase that would follow if the maximum Ft option (still unconfirmed) is adopted, a little over 40% can be offset by these three measures. Not all of it. Which is precisely why it is worth building the measurement foundation *before* the Ft is confirmed rather than reacting after the fact.

A note on the assumptions. All three figures are calculated independently from the same base case (Ft = 0.1623 THB/kWh, 800 kW, 250,000 kWh/month) and then simply added. In practice the measures interact: shifting load reduces on-peak coincidence and makes peak shaving easier, while cutting total consumption changes the kWh base for Lever 3. Treat these as an order-of-magnitude guide and re-run them against your own billing data before committing. Note too that the 4% in Lever 3 is an observed result from plants that kept an EnMS running, not something a system delivers automatically once installed. One further caveat on Lever 3: it is estimated at the effective rate of 4.14 THB/kWh, and because that rate has the demand charge baked into it, the figure runs slightly high for a measure that only removes kilowatt-hours. Building it up strictly from the on-peak and off-peak energy rates plus the Ft gives 72,000 × 4.1839 + 48,000 × 2.6037 + 120,000 × 0.1623 = 445,694 THB/year, some 51,000 THB below the 496,800 THB used above. Use the lower figure if you want the conservative view.

The cost of an energy monitoring system, in five layers

The question “what does an energy monitoring system cost?” cannot be answered until you fix the number of measurement points. What follows is a 30-point model, with cost broken into five layers. All figures are indicative and will move with site conditions, voltage, cable runs, and how difficult the required shutdowns are.

Rolled up:

Set against the model plant’s annual electricity cost of 12,418,488 THB, the year-one total corresponds to 620,000 ÷ 12,418,488 = approximately 5.0% at the low end and 1,860,000 ÷ 12,418,488 = approximately 15.0% at the high end (our calculation). From year two onwards, cost is dominated by the operations layer at 60,000–180,000 THB/year, so against annual savings of 1,035,564 THB the net benefit is 1,035,564 − 180,000 = 855,564 THB to 1,035,564 − 60,000 = 975,564 THB (if your contract keeps the platform-layer cloud subscription running, deduct that separately).

Five ways a cheap-looking quote gets expensive later

The most common mistake in comparing quotes is benchmarking only the unit price in the metering layer. What actually inflates the total, in most projects, is the implementation layer and the operations layer.

1. The cost of arranging shutdowns is not in the quote. Retrofitting CTs into main switchboards or motor control centres requires a power outage. If you cannot stop production, you are waiting for the annual inspection or a long holiday — and your installation date slides out by months. Beyond the electrical work itself, there is an invisible cost in the waiting.
2. Cloud pricing that scales with measurement points. Thirty points may fit comfortably in year one, but if the plan is to grow to 50 or 80 points next year, the platform layer can more than double against your year-one quote. Confirm whether pricing is per point or per site *before* signing.
3. Dashboard development treated as a change order. Standard screens often show nothing but total plant trend. The views you actually need in daily operations — energy intensity by equipment, shift-to-shift comparison, billed demand after ratchet — are frequently custom work. Document the requirements up front.
4. Overlooked communications and network work. If your buildings are separated, you need either fibre/LAN installation or a wireless relay. Long-range wireless such as LoRa cuts cabling, but in a structure full of metal you cannot judge feasibility without an on-site trial.
5. Zero hours budgeted for operations. The single largest cost is the time of the people who look at the data, chase the cause, act, and verify the result. A quote with nothing in the operations layer is, in effect, a quote that assumes you will never run the system.

For the investment decision itself, BOI privileges and the wider Thai investment environment sometimes need to be part of the picture. Our overview of factory automation in Thailand and BOI is a useful companion read.

Getting measurement points out of old plants and legacy equipment

“Our equipment is too old for this” is the objection we hear most often. The short answer is that age of equipment is not a barrier, because you are not modifying the machine — you are measuring on the electrical supply side.

Method 1: Retrofitted CTs (the most universal approach)

Inside the distribution board or motor control centre, a current transformer (CT) is fitted around the target circuit and wired to a power multimeter. Because this is independent of the equipment’s vintage, manufacturer, and control system, it works on machinery built in the 1980s.

• Strengths: works on anything; no modification to the equipment; captures real power, power factor, and current imbalance
• Weaknesses: requires a shutdown, because the work is inside a live panel. Even with split-core CTs, live working is not standard practice
• Practical notes: work requiring a main-incomer shutdown should be aligned with the annual inspection as a matter of course. Panel space (DIN rail, meter cut-outs) and whether the CT’s window will fit around the conductor both need an on-site check beforehand

Method 2: Capturing pulse and analogue outputs

Where existing kWh meters, flow meters, or gas meters already provide pulse or analogue outputs, those signals can be picked up by a data logger or IoT gateway.

• Strengths: in many cases no shutdown is needed; easy to extend to utility monitoring (compressed air, cooling water, steam, potable water, gas)
• Weaknesses: instruments without an output are out of scope; if you do not confirm the pulse weight from the instrument manual, your totals will be wrong
• Practical notes: watching electricity alone will cause you to miss compressor efficiency decay and air leaks. Looking at air flow and electrical power together makes root causes far easier to isolate

Method 3: Reading existing smart meters and incoming panel meters

If the revenue meter at the point of supply, or an existing digital power meter, supports Modbus or similar communications, you read directly from it.

• Strengths: no additional metering hardware, so this is the cheapest route; the fastest way to make whole-plant demand visible
• Weaknesses: you see one point only — the plant total — so you cannot isolate causes. You will know that today was expensive, but not why
• Practical notes: this is a perfectly good first step. Standing up whole-plant demand monitoring first, running the alarm process, then adding equipment-level metering incrementally is a realistic sequence

Treat the shutdown as a project milestone, not a detail

The biggest cause of delay for energy monitoring projects in Thai plants is not technology — it is arranging the shutdowns. Put the following into the schedule from day one:

• At the measurement-point-list stage, separate which panels require a shutdown and which do not
• Consolidate all shutdown-dependent work into the annual inspection or a long holiday (Songkran, New Year)
• Confirm lead times for any applications required by the industrial estate management company or PEA/MEA
• Agree the production impact of the shutdown window with production planning in advance
• Batch everything that must happen on shutdown day (CT installation, in-panel wiring, communications cabling all on the same date)

One more point: the platform that receives your power data can often be shared with equipment monitoring. If you are already running production monitoring, riding on the same gateways and the same network described in our guide to implementing factory IoT for equipment monitoring compresses the cost of both the collection layer and the implementation layer.

Choosing measurement points — never try to meter everything at once

Start with the big consumers

The classic failure mode is the reasoning “if we’re doing this, let’s measure everything.” The point count balloons to 100 or 200, the quote jumps, and the capital request stalls.

What you measure first is whatever eats the most electricity. The order varies by plant, but the usual suspects at the top of the list are:

Use a Pareto lens to scope phase one

The familiar rule of thumb is that the top 20% of causes account for 80% of the effect. The exact ratio differs from plant to plant, but for electricity it is generally true that a small number of large machines account for the bulk of total consumption, which makes covering that top group first the rational move.

The practical sequence is:

1. From the transformer capacity list and the single-line diagram (SLD), write out the capacity of each panel
2. From rated capacity and running hours, estimate approximate annual kWh for major equipment and sort descending
3. Working down from the top, take everything needed to cover the bulk of the total as your phase-one measurement scope
4. Cover the remainder with panel-level aggregate metering, and break it down to equipment level only when a question demands it

The 30 measurement points used in this article’s cost estimate reflect what a phase one scoped this way typically looks like. Standing up 30 points and expanding once the process is running delivers results faster than aiming at 100 points from the outset.

Extending to CO2 visibility

Once measurement is at equipment level, applying emission factors to consumption extends the same data into CO2 visibility — letting you answer supply-chain emissions enquiries from customers or headquarters at equipment or product level. That said, emission factors must come from official Thai published values, and this article does not deal with specific factors. Factor selection and boundary setting (the split between Scope 1, 2, and 3) should be fixed against the reporting requirements you face. As infrastructure, holding consumption by equipment and by time period means applying factors later is a straightforward addition.

Thailand’s Energy Conservation Promotion Act, Designated Factories, and the PRE requirement

Thailand imposes statutory energy management obligations on facilities above a certain size. An energy monitoring system also lowers the administrative burden of meeting them.

The governing law and the Designated Factory threshold

The governing legislation is the Energy Conservation Promotion Act B.E. 2535 (1992), administered by DEDE (the Department of Alternative Energy Development and Efficiency).

The thresholds for Designated Factory status are:

Meeting either criterion makes a facility a Designated Factory. Facilities are further categorised as 1,175–3,530 kVA and 3,530 kVA and above.

The point to watch is that the determination is based on the total capacity of electricity meters (kW rating, not kWh consumption) or transformers, not on actual contracted demand. Our model plant peaks at 800 kW, but if its transformers total 1,175 kVA or more it can still qualify. Because the determination depends on your installed configuration, verify it against the single-line diagram and the transformer nameplates. The assumption that “we’re only 800 kW, so this doesn’t apply to us” is exactly what produces a finding later.

The obligation to appoint an energy manager (PRE)

Designated Factories are required to appoint a PRE (Person Responsible for Energy).

The PRE executes energy management, records and reports data, and prepares energy conservation plans. In practice the role is usually held concurrently by someone in maintenance or production engineering — and that dual role becomes unworkable if daily data collection is manual. Where it takes several days at month-end to gather meter readings and transcribe them into Excel, producing the report becomes the objective in itself, and no time is left for the improvement work it is supposed to drive.

Alongside cost savings, automating the reporting burden is one of the larger benefits of an energy monitoring system. If data accumulates automatically, the PRE’s hours go into analysis and improvement rather than transcription. That value varies too much between plants to include in this article’s estimate, but it deserves explicit treatment in your business case.

ISO 50001 and energy management systems

The evidence behind 4% a year for a decade

ISO 50001 is the international standard for energy management systems (EnMS), with more than 38,000 certificates issued worldwide. Plants that have implemented an EnMS have been reported to sustain roughly 4% annual energy reduction for more than ten years (US DOE Better Plants).

What “4% a year for ten years” really tells you is that energy performance is not a one-off capital project — it is an operating cycle. The cycle ISO 50001 asks for is broadly:

1. Energy review (understanding where, on what, and how much energy is used)
2. Setting the energy baseline (EnB)
3. Setting energy performance indicators (EnPIs)
4. Establishing objectives and action plans
5. Implementation, monitoring, and measurement
6. Internal audit and management review, then the next cycle

Steps 1 and 5 do not exist without measurement data. Turn that around: once an energy monitoring system is in place and data is accumulating, the “measure, compare, record” portion — which is the bulk of what the standard requires — is handled structurally rather than by effort. Whether or not you intend to pursue certification, the framework is a sound template for designing your operating routine.

How to set your EnPIs

EnPI design is routinely overlooked. If you look only at total kWh, you will misread a drop caused by lower production volume as an energy saving. At minimum, put the following side by side:

• Total energy consumption (kWh/month)
• Energy intensity per unit of output (kWh/unit, kWh/tonne, etc.)
• Base load during non-production hours (the overnight and weekend floor)
• Billed demand (kW after the ratchet is applied)
• On-peak ratio (on-peak kWh ÷ total kWh)

Of these, the overnight and weekend base load is the one that pays out fastest with zero investment. Power flowing when nobody is producing anything is, by default, entirely a candidate for elimination.

A 90-day implementation roadmap

Getting an energy monitoring system live takes longer in internal alignment than in electrical work. Here are the stages that take you to operational status in 90 days, with the deliverable and the stakeholders for each.

Days 1–30: Establish the current state (no installation work yet)

This phase requires almost no additional spend.

What to do

• Collect every electricity invoice from the last 12–24 months
• Confirm your tariff class (Type 3 / Type 4), supply voltage, and contracted demand
• Tabulate monthly kWh, maximum demand in kW, and the on-peak/off-peak split
• Calculate the load factor: monthly kWh ÷ (maximum kW × 24 hours × days in month). For the model plant, 250,000 ÷ (800 × 24 × 30) = 43.4%
• Calculate the highest demand in the last 12 months × 70% (this is your ratchet floor)
• Gather the single-line diagram (SLD), the transformer capacity list, and rated capacities of major equipment

Deliverables: monthly energy register for the current state, baseline table, ratchet floor value, first-pass hypothesis on where savings sit

Stakeholders: finance/procurement (original invoices), the licensed electrical engineer and maintenance (SLD, transformers), production control (output data)

By the end of this phase you know what percentage of your bill is demand charge and what your on-peak ratio is. That is where the strategy branches: at about 10.3% demand charge, like the model plant, the energy-side measures lead; at 20–30%, peak shaving leads.

Days 31–50: Measurement design and quotations

What to do

• Build the measurement point list (one row per point: priority, method, shutdown required or not, panel location)
• Select the method for each point (retrofit CT / pulse output / read existing meter)
• Design the communications path (wired LAN, Modbus, wireless — verify inter-building distances and obstructions on site)
• Define KPIs (EnPIs) and document the dashboard screen requirements
• Draft demand alarm thresholds and a proposed shed priority list
• Identify all work requiring a shutdown and map it against the annual inspection and holiday calendar
• Obtain quotations broken out by the five layers (a single lump-sum figure cannot be compared)

Deliverables: measurement point table, cabling and communications design, screen requirements document, draft shutdown plan, layer-by-layer quotations

Stakeholders: production engineering, line supervisors (whether equipment can be stopped), IT (network, security), the electrical contractor, procurement

Days 51–75: Installation, configuration, and data validation

What to do

• Execute CT installation and in-panel wiring according to the shutdown plan
• Install gateways and loggers and confirm communications
• Reconcile against the invoice (do the system’s accumulated kWh and maximum demand agree with the billed values?)
• Publish the first dashboard release and review how it reads on the shop floor

Deliverables: measurement validation record (differences against the invoice and their causes), first dashboard release, as-built drawings

Stakeholders: maintenance, EHS (safety management for shutdown work), contractor, IT

The reconciliation step gets skipped more often than any other. Do not skip it. Reversed CT orientation, an incorrect CT ratio setting, and a wrong pulse weight are all common, and if you do not validate at the start, you will spend six months making decisions on wrong data.

Days 76–90: Embedding the routine (this is the real work)

What to do

• Define the escalation flow for demand alarms (whose device, within how many minutes, who shuts down what)
• Put the shed priority list on paper and post it on the shop floor
• Establish the rhythm: daily (yesterday’s peak and its cause), weekly (trend), monthly (reconciliation with the invoice and verification of results)
• Log improvement topics with an owner and a due date
• Agree the rules for measuring results (baseline comparison, how to normalise for production volume)

Deliverables: demand curtailment procedure, alarm escalation matrix, monthly report template, improvement topic register

Stakeholders: plant manager (decision authority), production managers and line supervisors (execution), the PRE, the overseas operations function at head office (results reporting)

If you want the shop-floor record-keeping to run cleanly as well, one option is to pair the system with an electronic forms tool such as i-Reporter by Cimtops Corporation, so that curtailment actions are recorded as they happen. Our own PEGASUS platform handles production management and energy management on a single foundation, which means energy intensity per unit produced is calculated without manual work. Whichever architecture you choose, whether the shop floor knows what to do when the alarm sounds is the single factor that separates success from failure.

Four common ways these projects fail

Failure 1: The dashboard becomes the deliverable

This is the most frequent one. A polished set of screens goes up on a large monitor, visitors are impressed — and three months later nobody is looking at it.

The root cause is that the screens were designed as something to look at rather than as a trigger for action. The fix is simple: for each screen, be able to write one sentence describing what the person looking at it decides and does. If you cannot write that sentence, do not build the screen. Alongside that, define as an actual job who checks what on a daily, weekly, and monthly basis, and make the outcome of those checks a written record.

Failure 2: Too many measurement points, so the project dies

“If we’re going to do it, let’s do all of it” produces a 200-point quotation, and the capital request stops the moment someone reads the total. Or the volume of installation work makes the shutdown impossible to schedule and the start date slips a year.

The remedy, as above, is to scope phase one at around 30 points focused on the large consumers. Standing up small and fast, demonstrating a result, and using that result to justify phase two funding covers the whole plant sooner than trying to do it all at once. Just make sure, at the outset, that your gateways have spare ports and that you know the point ceiling in your cloud contract.

Failure 3: The demand alarm fires and nobody has a procedure

The alarm is sounding. But the shop floor does not know what may be shut down. It might affect quality. It might create problems downstream. Nobody is sure they have the authority. So nothing happens, and the peak is recorded.

The remedy is to decide before the alarm ever sounds:

• The curtailment priority order (least impact on quality, safety, and delivery first)
• How long each item may be stopped, and whose call it is
• The restart procedure and who downstream must be notified
• Alarm timing (firing after the 15-minute interval has closed is too late — use a predictive alarm within the interval)

One page is enough for this procedure. In fact, a procedure that does not fit on one page will not be used.

Failure 4: Not knowing about the ratchet, and getting the business case wrong

Someone calculates “we can cut 300 kW, so 132.93 × 300 = 39,879 THB/month”, gets the capital approved, and then the invoice falls by only 31,903 THB/month. The 7,976 THB/month difference (our calculation) is the portion blocked by the 560 kW ratchet floor.

The absolute amount may look small, but “the projection did not match the result” is the fact that follows you into your next IT capital request. At the estimating stage, always check the highest demand in the last 12 months × 70% and build it into the payback calculation. Equally, explain up front that the benefit grows once the ratchet floor resets after 12 months — then nobody has to reinterpret the numbers later.

Frequently asked questions

What is an energy monitoring system?

It is a system that automatically captures how a plant uses electricity and other energy, makes that usage visible by equipment and by time period, and connects it to concrete reduction measures. It is also referred to as a power monitoring system, an EMS, or an energy management system.

It consists of five layers: (1) the metering layer (power meters with CTs, pulse-output instruments, and so on); (2) the collection layer (IoT gateways, data loggers); (3) the platform layer (visualisation software, cloud); (4) the implementation layer (electrical work, configuration, screen development); and (5) the operations layer (maintenance, running the improvement cycle). Scope can extend beyond electricity to utility monitoring — compressed air, cooling water, steam, and gas.

The important qualification is that the system’s job stops at measuring and displaying. What actually reduces the bill is the operating routine built on that data: demand peak shaving, load shifting, and continuous energy reduction. If the question you are really asking is how many measurement points you need and what they will cost, our companion article on measurement point design and cost breakdown for a power monitoring system goes into that in detail.

How much does an energy monitoring system cost?

For a 30-point model, the initial cost (metering + collection + implementation layers) is 500,000–1,440,000 THB, and the year-one total (layers 1–5, including platform and operations) is 620,000–1,860,000 THB. Both figures move with site conditions.

The component guide figures are 8,000–20,000 THB per metering point, 30,000–80,000 THB per gateway, 5,000–20,000 THB/month for visualisation software or cloud, 200,000–600,000 THB for electrical work, configuration, and screen development, and 60,000–180,000 THB per year for maintenance and operations.

For the model plant in this article (250,000 kWh/month, 800 kW maximum demand), estimated annual savings are 1,035,564 THB, giving a payback of 620,000 ÷ 1,035,564 = 0.60 years to 1,860,000 ÷ 1,035,564 = 1.80 years — that is, approximately 0.6 to 1.8 years. When comparing quotations, insist on the five-layer breakdown. A lump-sum figure tells you neither what is expensive nor what has been left out.

Will visualising power consumption reduce the bill on its own?

No. More precisely, visibility itself produces no saving. Savings occur only when what you see causes you to change how you operate.

That said, it is equally true that without visibility you cannot select a measure at all. If you do not know which equipment consumes how much during on-peak hours, you cannot choose what to shift; if you do not know when peaks occur, there is nothing to shave. Visibility is a necessary condition, not a sufficient one.

In practice, results require all three of the following together:

1. Numbers (measurement data by equipment and by time period)
2. Decision criteria (thresholds, priorities, KPIs)
3. Action (a procedure and the authority defining who does what, and when)

Deploying only item 1 while items 2 and 3 remain undefined is precisely the “dashboard becomes the deliverable” failure described above.

What should demand monitoring actually display?

These five things:

Items three and four in particular are not standard on most out-of-the-box dashboards. Unless you state them explicitly as requirements, you will be handed a screen that shows actual kW and nothing else.

Equally important is being able to trace back the log by equipment for the date and time a peak occurred. Only when you can resolve it to the level of “last Monday between 10 and 11, the compressor and chiller start-ups coincided with the furnace ramp-up” does it translate into a specific countermeasure such as staggering start times.

Can old equipment be measured?

Yes. You are not modifying the machine itself; you are measuring on the circuit that supplies it. There are three approaches:

1. Retrofit CTs: fit a CT to the target circuit inside the distribution board or MCC. The most universal method, independent of equipment age or manufacturer. Requires a shutdown, because the work is inside a live panel
2. Capture pulse or analogue outputs: if existing kWh meters or flow meters have output terminals, take the signal from there. Often possible without a shutdown
3. Read existing smart meters or digital power meters: if communications are supported, read the value at the point of supply. The cheapest option, but it shows only one point for the whole plant

The real difficulty is not technical — it is arranging the shutdown. CT installation requires an outage, so plan from the outset to consolidate the work into the annual inspection or a long holiday (Songkran, New Year). Leave that out of the schedule and you will spend months with the hardware sitting in a box and no date to install it.

Summary

The key points from this article:

• Thailand’s average tariff was 3.88 THB/kWh in January–April 2026 and 3.95 THB/kWh in May–August. For September–December, four options with an Ft of 0.1623–0.9482 THB/kWh went to public hearing, but no decision had been made at the time of writing
• The industrial tariff decomposes into a 3.78 THB/kWh base tariff + Ft + TOU energy charges + demand charge. The base tariff and the Ft are outside a plant’s control; what a plant can change is three things — when it consumes (TOU), how high it lets its peak go (demand), and how many kWh it consumes at all
• The demand charge is set by a single maximum 15-minute average in the month, and is subject to the ratchet: billed demand = MAX(this month’s actual, highest of the last 12 months × 70%). You cannot make an investment decision without knowing both rules
• The model plant (250,000 kWh/month, 800 kW) pays 1,034,874 THB/month at an effective rate of 4.14 THB/kWh, with the demand charge at approximately 10.3% — because its load factor is 43.4%. The 20–30% rule of thumb describes low-load-factor plants
• If the Ft goes to 0.9482, that is 196,475 THB/month and 2,357,700 THB/year in additional cost (unconfirmed scenario)
• Three measures (peak shaving 159,516 + load shifting 379,248 + 4%/year reduction 496,800) total 1,035,564 THB/year, or approximately 8.3% of the annual electricity bill
• Cost for 30 measurement points is 620,000–1,860,000 THB in year one, with payback of approximately 0.6–1.8 years
• Old equipment can be measured via retrofit CTs, pulse outputs, or reading existing meters. The hard part is scheduling the shutdown
• If transformers total 1,175 kVA or more (among other criteria), Designated Factory status triggers the PRE appointment requirement. Automating measurement directly reduces the reporting workload
• Four failure patterns: the dashboard becomes the deliverable; too many points to survive approval; no procedure after the alarm; and misjudging the business case through ignorance of the ratchet

There is one thing you can do right now. Line up the last 12 invoices, find your highest demand, and multiply it by 0.7. That number is the floor your electricity bill will not go below.

Even at the stage of taking the invoice apart and drawing up a candidate list of measurement points, an outside perspective tends to sharpen the questions quickly. TOMAS TECH works with plants across Thailand on first-pass diagnostics — where the savings sit in your current invoice — and on scoping measurement points around the condition of your existing switchgear. There is no expectation that a project follows; a conversation at the stocktaking stage is entirely fine. You are welcome to get in touch through our contact page.

References

• ERC May–August tariff at 3.95 THB/kWh (Nation Thailand)
• Proposed September–December Ft (Thailand Construction)
• Demand charge, TOU, TOD, and ratchet explained (CapSolar)
• Factory electricity cost and effective rates in Thailand (CapSolar)
• How the Ft (fuel adjustment charge) works (CapSolar)
• Energy Conservation Promotion Act B.E. 2535, English translation (EPPO)
• Designated Factory criteria and PRE requirements (APEC EGEEC Thailand update)
• ISO 50001 and 4% annual reduction (US DOE Better Plants / Better Buildings Solution Center)
• ERC public hearing (Bangkok Post)