Energy Monitoring System: Cut Factory Power Costs in ASEAN

Your electricity bill keeps rising, yet nobody can say which machine is spending the money

One of the most common requests reaching us from plants in Thailand and Vietnam is “our electricity cost is going up, please help”. The obvious follow-up — how much does each major asset consume per month? — almost never gets an answer. The plant holds one utility invoice and one aggregated figure for the whole site: the amount paid is known, what generated it is not. The first thing an energy monitoring system fixes is that disconnect between the invoice and the shop floor.

Output is counted daily and defect rates tracked by process, yet electricity surfaces once a month as a bill that lands in finance. Any raw material would have goods receipt, issue and stock records by item code — yet electricity — a raw material bought in bulk every single month — is consumed with no item-level record at all.

Cost is no longer the only pressure. Group sustainability asks for CO2 emissions by site, a customer exporting to Europe asks for emissions data per product, and banks, the regional HQ and investment committees all have a line for energy intensity. Being unable to answer with numbers has shifted from an internal inconvenience to a credibility risk.

There is also a trap specific to this region. The energy-saving playbook engineers bring from a parent plant is built around reducing total kWh. Industrial electricity in Thailand is billed through time-of-use (TOU) pricing, where the unit rate changes sharply by time of day, plus a demand charge based on the highest demand recorded in the month. When you use power, and how hard you ramp up, feed straight into the bill. In Vietnam a pilot of a two-part tariff separating capacity from consumption is getting under way. The same kilowatt-hour costs a different amount depending on the hour — yet far too many improvement projects here stop at “we changed the lighting to LED”.

This article defines what an energy monitoring system covers, then sets out how to decide what to measure, where to start, what drives the investment and who to ask.

What an energy monitoring system actually is: EMS, power monitoring and demand monitoring

Every vendor uses different words, and the same word carries different meanings from one proposal to the next. Collecting three quotations and finding they do not describe the same thing is ordinary.

By function, the system resolves into four layers, each assuming the one below it is built correctly.

“Power monitoring system” usually means layers 1 and 2 only. “EMS” normally adds layer 3, and in a manufacturing context is sometimes called FEMS (factory energy management system). “Demand monitoring” and “demand control” sit in layer 4 — raising an alarm, or dropping load, when contracted demand is about to be exceeded.

How far up the stack a proposal reaches changes the price, the organisation required and the benefit obtained. Layer 2 is meters, communications and a dashboard. From layer 3 upward you must join production data such as output and running hours, which forces a design decision on the denominator of your intensity metric. Layer 4 throttles equipment, so it cannot exist without agreement between production and maintenance.

Watch also for the confusion between EMS and BEMS, its building-oriented equivalent. A factory mixes building-type loads such as HVAC and lighting with manufacturing loads. Deploy a building product unchanged and you can optimise comfort but never derive energy per unit of product, because nothing links the data to production records. What a factory needs is a system that knows about production, which is why it belongs alongside your MES strategy — see how to approach production management system (MES) adoption in a Thai factory.

This is not pedantry. It is normal for supplier A to quote layers 1 and 2, supplier B layers 1 to 3 and supplier C all four; take the cheapest and you later hear “analysis is a separate licence”. Decide how far up the stack you are buying before you issue the enquiry.

Breaking down the electricity tariff structure of a Thai factory

To reduce a Thai factory’s electricity cost you first have to take the invoice apart. Without that structure there is no way to separate measures that work from measures that do not.

For the May–August 2026 period the average tariff in Thailand is reported at 3.95 baht/kWh, up about 1.8% from 3.88 baht/kWh in the preceding January–April 2026 period (Thailand-specific figures reported by Wise Digital). The Ft for the same period is 0.1623 baht/kWh, set by the ERC — roughly 4.1% of that average rate.

The Ft is revised every four months, so the rate can move three times a year for reasons unrelated to your operations. Compare this month with the same month last year and you cannot say how much of the gap came from your own behaviour and how much from an Ft revision. Separating internal from external factors is one of the underrated reasons to meter.

Published reference values for the industrial TOU tariff put on-peak at 4.1025 baht per unit (Monday to Friday, 09:00–22:00) and off-peak at 2.5849 baht per unit (22:00–09:00 plus weekends and public holidays), with Ft and 7% VAT added on top (Thailand-specific reference values). The gap is 1.5176 baht per unit, a ratio of about 1.59 times: the same kilowatt-hour costs nearly 60% more during a weekday daytime hour than overnight or at the weekend.

Reference values circulating for large factories (Type 4) show on-peak at 5.27 baht/kWh and off-peak at 3.80 baht/kWh (Thailand-specific reference values), a ratio of about 1.39 times. These two sets are very likely not on the same basis: whether a quoted rate is the energy charge alone or the figure after Ft and other additions, and which voltage class and contract type applies, changes the number. Because different sources take the breakdown differently, always verify against the itemised detail on your own invoice. Rates differ across high, medium and low voltage classes and by contract type, and the Ft moves every four months. The figures here were the published values at the time of writing; reconcile them against the latest ERC publication and your own bill. Whether the 3.95 baht/kWh average includes VAT is also treated differently by different sources.

Then the item most often overlooked by engineers arriving from elsewhere: the demand charge. It is billed on the maximum demand within the month, generally the highest average power over a fixed interval — commonly fifteen minutes. That settlement interval can differ by contract type and by distribution utility, so confirm it against your own supply contract or invoice. It sits alongside, not inside, the consumption charge, and unit rates vary by voltage class and contract type.

How it is determined matters. A plant running 24 hours a day, 30 days a month has 720 operating hours, or 2,880 fifteen-minute intervals, and exactly one of them sets the demand charge — about 0.03% of the month. However carefully you conserve through the other 99.97% of the time, one quarter-hour in which several machines started together decides that month’s charge. This is a structural reason why consumption reduction alone fails to move the bill, and the invoice will never tell you when that interval occurred: identifying which assets ramped up together requires measured time-series data.

Why cutting total kWh alone does not cut your electricity bill

There are three ways to reduce a factory’s electricity cost. Absolute reduction improves equipment efficiency, cuts standby consumption and repairs leaks; this is where most conventional energy-saving activity concentrates. Peak shifting keeps total consumption the same but moves it from on-peak to off-peak, which under TOU saves the rate difference. Peak shaving and demand control suppress maximum demand: do not start things together, stagger start-up sequences, forecast demand and throttle before the ceiling.

Many engineers arrive with only the first tool. Under the Thai TOU structure the second and third are disproportionately powerful.

Assume a plant consuming 500,000 kWh a month — 300,000 kWh on-peak, 200,000 kWh off-peak — priced at the industrial TOU reference values above.

• On-peak: 300,000 × 4.1025 = 1,230,750 baht
• Off-peak: 200,000 × 2.5849 = 516,980 baht
• Total: 1,747,730 baht

Now shift 30,000 kWh, or 10% of on-peak consumption, into the off-peak window. Not one unit of output is given up; only the timing changes.

• On-peak: 270,000 × 4.1025 = 1,107,675 baht
• Off-peak: 230,000 × 2.5849 = 594,527 baht
• Total: 1,702,202 baht

That is 45,528 baht a month, or 546,336 baht a year — about 2.6% of the original energy charge. Note that this compares energy charge only: it excludes Ft, VAT and the demand charge, and the demand charge does not move in this calculation, so the reduction against the total invoice would be smaller than 2.6%.

To capture the same amount through absolute reduction: the blended rate is 1,747,730 ÷ 500,000 = about 3.495 baht/kWh, so removing 45,528 baht means eliminating about 13,025 kWh, or 2.6% of total consumption. The two figures both landing near 2.6% is not a coincidence — it follows from converting the money back into kWh at the blended rate. What differs is not the percentage but the work required to get there. Absolute reduction requires real equipment or operational change, usually with capital attached, whereas peak shifting can reach a comparable result by re-sequencing the plan.

Where you save also changes what you earn. Removing 30,000 kWh is worth 123,075 baht from the on-peak window and 77,547 baht from off-peak — same kilowatt-hours, a 1.59-times difference in money. Stopping a compressor idling during the day is worth substantially more than stopping the same idling at night. Where the on-peak spread is narrower and total-volume reduction dominates improvement activity — Japan-specific practice is the clearest example — this ranking is rarely front of mind. In Thailand it is unmissable.

What can realistically be shifted? Re-sequencing production, moving batch processes to night shift, pre-cooling or thermal storage on chillers and HVAC, compressed-air storage, retiming washing and drying, and changing when AGVs and forklifts charge. Battery charging is a persistent blind spot: if everything is plugged in at the lunch break, that habit inflates the on-peak profile. Plants investing in automated transport should treat charging schedules as an energy question too — see the issues to work through when introducing AGVs and AMRs.

Deciding when things run is the planning department’s job, so if “this window is expensive” cannot reach next week’s schedule, no improvement occurs. Our comparison of production schedulers and how to select one shows how an electricity constraint can be built into the plan.

Peak shaving demands more discipline again. Because demand is set by the maximum fifteen-minute average, noticing after the interval closes is useless: the practical implementation predicts part-way through an interval that continuing at the current rate will breach contracted demand at the boundary, then alarms or curtails low-priority load. Demand alarm devices are widely deployed in Japan (Japan-specific practice) but frequently absent in Thai and Vietnamese plants, which makes it unusually accessible headroom. If you introduce automatic shedding, agree in advance with production what may be dropped — an operating-rules problem, not a technology problem.

Vietnam sites: what changes when the two-part tariff pilot begins

The average electricity price of EVN, the state utility, rose 4.8% with effect from 10 May 2025 to 2,204.0655 dong/kWh excluding VAT (Vietnam-specific figure). Price increases are a shared regional trend; what makes Vietnam interesting is that the structure is changing.

The Ministry of Industry and Trade (MOIT) has approved a pilot of a two-part tariff, separating a capacity charge paid against contracted capacity from an energy charge paid against actual consumption — conceptually close to the Thai combination of demand plus energy charge. Scope is industrial customers supplied directly by EVN with average monthly consumption of 200,000 kWh or more (Vietnam-specific threshold), which over 30 days is roughly 6,667 kWh a day, so any medium-sized plant or larger is inside the frame. Full introduction is targeted from 2026, with the operational pilot stated as July 2026 to July 2027 (Vietnam-specific timeline). The TOU time bands were separately restructured under Decision 963/QĐ-BCT dated 22 April 2026 (Vietnam-specific regulation) — and when bands are redrawn, the window your team treats as “the cheap hours” may no longer be cheap.

Two consequences follow. Sites that only ever watched total consumption acquire a new control variable, maximum demand. And setting a contracted capacity requires you to justify how much capacity you genuinely need: contract too much and you pay a fixed cost indefinitely, too little and operations suffer. That judgement needs a measured history of how demand varies within the month and across the year — and if you start measuring after the rules take effect, the history does not exist yet.

Beyond the above, this article avoids asserting details of eligibility, conditions or incentives; the Vietnamese framework is in transition. For any decision with commercial consequences, obtain written confirmation from the competent authority or a qualified local adviser.

Two countries, two design philosophies for the tariff

For groups operating on both sides of the border, the difference changes what good energy management means at each site.

Thailand has a mature structure in which time-of-use pricing and a demand charge have coexisted for a long time. The price signal targets two behaviours at once: move consumption out of the expensive window, and flatten the peak. Improve the load factor and you are rewarded twice, on next month’s invoice.

Vietnam has largely priced industrial supply through consumption with time bands, and the two-part pilot moves toward pricing contracted capacity as a distinct product. A capacity element rewards accurate forecasting and honest sizing of the connection more than minute-by-minute peak suppression inside a billing period. Both will matter as the pilot matures, but the first-order discipline it creates is capacity planning.

So the same measure carries a different payback in each country, and the same KPI can be meaningful in one plant and meaningless in the other. “Reduce kWh by 5%” is reasonable where consumption dominates the bill and poor where a capacity or demand element is the growing part of the cost. Set the target at group level; let the measure be chosen locally against the local tariff.

One preparation is certainly not wasted whatever the rules become: hold a full year of your own load curve at fifteen-minute granularity. Under every conceivable tariff, the distribution of your demand across time is decision-relevant, so regulatory uncertainty is not a reason to postpone measurement.

What to measure and how deep: three metering tiers and how to prioritise

The commonest failure in these projects is metering the incomer and stopping.

Stop at tier 1 and you get a monthly impression: more than last month, less than last month. It does not convert into an action. Meter every asset at tier 3 from the outset and you reach several hundred points and a capital request that never gets approved. Fix tiers 1 and 2 first, then extend tier 3 selectively to the assets that typically sit at the top of a factory’s consumption profile.

Compressors (compressed air). Often the largest consumer and the one with the most headroom. Air leaks, discharge pressure set higher than necessary and idling through breaks and night shifts show up immediately once measured, and whether multi-machine sequencing control works as intended cannot be judged without it. Top of the list.

Chillers and refrigeration. Heavy consumers, strongly correlated with ambient temperature, so pairing consumption with weather data is productive, and the leading candidate for peak shifting via thermal storage. In the Thai climate they run all year, so improvements persist across twelve months rather than a season.

HVAC. In plants with cleanrooms or temperature- and humidity-controlled areas, air conditioning can rival production equipment. Set points, fresh-air volumes and filter loading make consumption move on operational decisions.

Furnaces and heating equipment. Batch-operated heating is a prime suspect for demand peaks through inrush at start-up, which makes it the primary target for staggered starts.

Large motors (pumps, blowers, conveying). Individually mid-sized, collectively significant because of the number of units. Verifying the benefit of variable-speed retrofits also requires measurement.

The decisive design question is what you will compare against. Only with production output, running hours, ambient temperature or shift count in the denominator do you get a comparable intensity metric, and where that denominator comes from depends on production-side data structures. Defer it and you will finish the installation only to find you cannot compare.

Making energy data comparable across countries

Groups with plants in more than one country hit an extra problem here, worth designing for before the first meter is fitted.

Currency is the obvious trap. THB and VND cannot be added, and converting introduces an exchange-rate effect unrelated to plant performance. Keep physical units — kWh, and energy per unit of the denominator — as the comparable basis across sites, and treat cost as a local view rather than something to aggregate without a stated conversion policy.

The denominator is subtler. If one plant reports kWh per finished unit, another kWh per tonne of input and a third kWh per direct labour hour, the table looks comparable and is not. Agree a small set at group level — typically one physical output denominator and one time denominator — and require every site to publish both, whatever local metric it also uses.

Then the boundary definitions that quietly diverge. Does the site figure include the canteen, the dormitory, the wastewater plant, on-site solar generation? Is consumption reported for the calendar month or the utility’s billing cycle? Is cost inclusive or exclusive of VAT? Are on-peak hours defined by the local tariff, so that the same label means different clock hours in different countries, or by a group-standard window that maps to nobody’s bill? None of these questions is hard, and all are hard to fix retrospectively, because history cannot be re-collected under a definition that was not applied at the time.

Network design and security cannot be deferred either. It is common for a gateway to end up with one foot in the control network and one in the office network — convenient, and a real concern for segmentation. Settle the separation policy and the route to any cloud connection at design stage, informed by a practical approach to OT security in the factory. Changing it later is far more painful than adding meters.

For context, a global estimate suggests mid-sized manufacturing plants waste in the order of 15–25% of their energy through undetected inefficiencies. Its monetary conversion is presented in US dollars, so do not map it directly onto your site. Designing your metering plan on the assumption that roughly a fifth of consumption is invisible is nevertheless defensible: what you cannot see, you cannot improve.

Where the money goes: reading a quotation in five cost layers

Absolute figures vary enormously with plant size, number of points, building layout and the condition of existing equipment, so this article quotes no prices — only what drives them.

Layers 1 and 2 are specified most concretely and are easiest to compare, being unit price times quantity. CT selection must match the current rating of the circuit, and “buy a lot of cheap CTs” is not an accuracy strategy: a directional trend, a basis for allocating cost between departments, and primary data for an emissions report demand different confidence levels.

Layer 3 is where capex and running cost blur. A low-capex offering charging per point per month and a higher-capex offering with a flat per-site fee can swap places over five years, so if you plan to add points later, check the slope of the per-point charge. Check retention too: a contract keeping only thirteen months of data makes year-on-year intensity comparison structurally impossible.

Layer 4 is the most consistently underestimated line. Fitting CTs generally requires the circuit to be de-energised, which in a 24-hour plant is the binding constraint on the whole project; if the only opportunity is the annual maintenance shutdown, missing it means waiting another year. Panels with no spare space add modification cost and lead time.

Then layer 5, operating effort. Because it never appears on a quotation it is treated as zero in the capital request and becomes visible only after go-live. An energy monitoring system is not a device that reduces your bill by being installed: money moves only when somebody reads the data, forms a hypothesis, implements a countermeasure and verifies the effect. Reviewing the monthly report, chasing anomalies and verifying results is not realistic as a side activity. Appoint someone, formally allocate a share of an engineer’s time, or contract operational support externally — and decide while the investment is still being approved.

One note on labour cost in Thailand: the minimum wage is set by province and expressed as a daily rate, in a range of 337–400 baht per day (Thailand-specific figures). It is daily, not hourly, and not uniform nationally. Someone who can interpret energy data and drive improvement is not a minimum-wage role, so those figures cannot be used to cost layer 5; what you need is time from an experienced engineer who understands the equipment and can negotiate with the shop floor.

On payback, some vendor-published material reports electricity cost reductions of 12–18% and payback in 12–18 months from IoT submetering. Those are vendor-published global indications, not measured results from Thailand or Vietnam. Real payback depends on the size of your bill, your on-peak share, the headroom in your demand profile and whether you can execute improvements, so build your own estimate from your own invoices.

Making a retrofit work in an existing plant

A greenfield project designs metering in from the start. Almost every real enquiry concerns retrofitting a plant that is already running.

Split-core CT retrofit. The standard method. A split-core (clamp-type) CT needs no cable cutting, so installation is comparatively quick — but because the work is close to live parts, de-energising the circuit is normally a precondition. You also need panel space for the CT and a route for the secondary wiring.

Pulse outputs. Where an existing energy meter has a pulse output terminal, taking that signal gives consumption data without a new CT. Inexpensive, but what you get is a pulse train representing cumulative energy: no power factor, no current waveform. Adequate for trends, thin for detailed analysis.

Reading existing meters over Modbus/RS-485. If digital power meters are already installed, polling them over Modbus RTU (RS-485) is usually the highest-value option available. Watch cable routing, the address plan, and physical-layer settings such as baud rate and termination resistors; where installed meters span several manufacturers and generations, assembling register maps takes real effort.

Taking data from existing PLCs. Machine control PLCs sometimes already read power values or analogue inputs, and retrieving those over Ethernet avoids new meters. But accessing a running PLC means touching the control system: restrict access to read-only and verify the impact on scan time first. Because that judgement presumes an understanding of the plant’s automation architecture, review it against the wider picture of factory automation adoption in Thailand.

Non-contact and temporary measurement. Portable clamp meters and data loggers can measure for a few weeks to establish a trend and narrow the permanent plan. Used as a diagnostic, this removes a lot of wasted investment.

The largest obstacle is not technical but scheduling — specifically the plant that cannot be shut down. Work in three steps. First, take everything that needs no shutdown: Modbus polling, pulse outputs and portable logging can generally be done with equipment running. Second, list the circuits that genuinely require new CTs and execute them together during the annual shutdown; because the value depends on how much you complete in one window, panel surveys, procurement and method statements must be finished well in advance. Discovering on the day that there is no spare space in the panel is the worst available outcome. Third, for high-priority circuits that cannot wait, meter an upstream feeder and infer by difference, or apportion using equipment run signals — accuracy suffers, but it beats “we cannot measure it, so we will do nothing”.

The shutdown opportunity is the rate-limiting step for the whole project, governing the schedule far more than the software build. Stating that in the paper you put to management makes every later scheduling conversation easier.

CO2 visualisation, CBAM and reporting to the parent company

The factor with the fastest-growing influence on these investment decisions is the demand for emissions reporting.

Structurally it is simple. Multiply measured purchased electricity in kWh by the applicable emission factor and you have indirect emissions from electricity — the category generally classified as Scope 2. The data your monitoring system already collects is the primary data for emissions reporting. A plant with no metering instead picks the total off the invoice by hand every cycle and applies factors in a spreadsheet; the moment anyone asks for a breakdown by site or product, that method collapses.

The external driver is the EU’s Carbon Border Adjustment Mechanism (CBAM). CBAM began with a transitional period of reporting only from October 2023 and moved into its definitive period on 1 January 2026 (EU regulation). Authorised declarants importing more than 50 tonnes a year of CBAM-covered goods incur an obligation to surrender CBAM certificates in line with embedded emissions. From 1 January 2026 reporting moves to an annual declaration and reported emissions are subject to verification, with the first declaration deadline on 30 September 2027 (EU regulation).

The critical detail is the scope of embedded emissions. Alongside direct emissions from the production process, there are goods for which indirect emissions arising from electricity consumption fall within scope. This does not apply uniformly to all covered goods — treatment varies by product — so check how your own products are handled. If indirect emissions are in scope for what you make, the value of measured electricity data rises steeply, because what is required is not a site total but data granular enough to allocate to a product.

For scale, direct financial exposure to CBAM across Vietnam’s four carbon-intensive export sectors has been estimated at roughly EUR 580 million a year (Vietnam-specific estimate). Vietnam’s 2024 steel exports to the EU exceeded EUR 2.2 billion, against approximately EUR 120 million for Thailand (Vietnam- and Thailand-specific trade figures) — on this trade measure Vietnam’s EU steel export scale is roughly eighteen times Thailand’s, so the two countries are in materially different positions.

It is worth resisting over-dramatisation: CBAM concerns exports into the EU and your plant is not automatically in scope. The route that matters more widely is indirect — even if you do not export to the EU, a customer who does may ask you for emissions data. In automotive components, electronics and metal processing, enquiries of this shape are already increasing.

A further route is the reporting request from the parent company or regional HQ. Group sustainability consolidating emissions asks overseas sites for monthly or quarterly submissions, and without a measurement base somebody assembles those numbers by hand every time — which brings key-person dependency and errors, both indefensible under audit or third-party verification. Reframing the system as the primary data platform for reporting, rather than only a cost-reduction device, often clears internal approval more easily.

The choice of emission factor (utility-published values versus national grid factors) and the required reporting format depend on both the regulatory framework and group policy, so confirm them in writing with the competent authority or a qualified local adviser.

ISO 50001 and how to use the available frameworks

The framework that turns energy management into a durable system is ISO 50001, the international standard for energy management systems.

One study found companies holding ISO 50001 certification reduced energy intensity by approximately 26%; this is research based on micro-data from Chinese manufacturing and is not a figure for Thailand or Vietnam (China-based study). Separately, plants implementing ISO 50001 have been reported to achieve 12–18% energy consumption reduction in the first year and to sustain around 4% annual reduction for more than ten years (global reporting). Sustaining 4% a year for a decade compounds to 0.96 to the power of ten, about 0.665 — a cumulative reduction of roughly 33%. Time to reach a certifiable state is generally put at 9–18 months (global benchmark).

There is no magic in the standard itself, but there is genuine effect in the processes it demands: an energy baseline, defined energy performance indicators, objectives with action plans, and continual review. That is the activity described throughout this article, fixed in documented form, and adopting the skeleton as an internal rule is worthwhile even without certification.

So, certify or not? Three tests. Is certification demanded by a customer or the parent company? Does the group need a common management framework, which makes cross-site comparison far easier once you have several sites? Can you carry the effort of achieving and maintaining it? If none applies, run the system without certification and pursue it when it becomes necessary.

Now the Thai framework. Under the Energy Conservation Promotion Act, enacted in 1992, designated factories and designated buildings must record and report energy consumption and prepare energy conservation plans for submission to DEDE, the Department of Alternative Energy Development and Efficiency (Thailand-specific legislation). Sources indicate a submission frequency of once every three years, but administration can change, so whether your site is a designated facility, and the current cycle and format, should be confirmed in writing with the competent authority or a qualified local adviser.

Complying with this obligation and deploying an energy monitoring system require almost the same work. As a compliance chore, assembling the data is a recurring burden; with a measurement base in place, the report is an extraction from something you already hold.

On the investment side, capital expenditure on approved energy-saving equipment qualifies for a 150% deduction under an existing framework (Thailand-specific incentive). The Board of Investment (BOI) also offers incentives for energy conservation, renewable energy and environmental impact reduction, including exemption or reduction of import duty on machinery and corporate income tax exemption of up to eight years (Thailand-specific incentive).

BOI incentives are widely misunderstood, so be precise. Structures differ by category. Under the Smart and Sustainable Industry measures, corporate income tax exemption is 50% as the baseline; it reaches 100% only where automation or robotics are introduced into the production line and at least 30% of the value of the replacement machinery is sourced from Thailand’s domestic automation industry (Thailand-specific conditions). There is no unconditional 100% exemption via BOI, and an investment plan built on that assumption will not survive contact with the application.

The current BOI incentive cycle is treated as reaching a milestone at the end of 2026, and a review of conditions is expected across 2026 and 2027. Framing a long-horizon plan on the assumption that today’s conditions are permanent is unwise. If incentives form part of your case, obtain written confirmation from a ที่ปรึกษาด้านภาษี (tax consultant) or ที่ปรึกษากฎหมาย (legal consultant); eligibility depends on the timing of the application, the business category and the equipment specification.

Five failure patterns in energy monitoring projects

Five failure modes recur. None is technical; all are failures of design and operation.

1. Building a dashboard and stopping there. An impressive set of graphs goes live, and three months later nobody opens it, because nobody defined who looks at the graph, when, and what decision follows. A dashboard is an input to a decision, not a deliverable. At design stage, specify the routine concretely — “the on-peak consumption share for the previous week is presented at the Monday production meeting”. Embedding it in an existing forum drives results far more than the visual quality of the screen.

2. Metering only the incomer and never finding the cause. You will know demand rose this month; you will not know which asset caused it. The discussion lands on “everybody please save energy”, and the same thing happens next month. Narrowing down the culprit becomes possible only at layer 2, feeder and area level. Budget rarely stretches to every asset, but it should always stretch to feeder level.

3. No named owner of improvement. Energy cuts across every department, so it becomes everybody’s issue and therefore nobody’s. Production prioritises output, maintenance prioritises stability, finance watches cost; nobody is lying and nothing improves. Name a person accountable for energy intensity and give them decision rights and a forum. It is an organisational problem rather than a system problem, and nevertheless the largest single determinant of success.

4. Reporting to the group format and nothing else, so the shop floor never uses it. When filling in a mandated template becomes the purpose, the exercise is pure added workload for the plant. Design two things separately: the views the site uses for its own improvement — daily trends by asset, comparison by shift, alerts on anomalies — and the aggregation used for group reporting. The first is what makes the second trustworthy, and the order is reversed at a great many plants.

There is a sharper version in multi-country groups. A regional HQ standardises the reporting format on the metric that made sense where the format was written, very often total kWh and kWh per unit of output, and the site then manages what it reports. Where the local cost driver is peak demand or a capacity charge, the site can hit the group consumption target while its bill goes the wrong way, with the reporting pack green throughout. The remedy is not to abandon standardisation but to require every site to report one locally-chosen cost-driver metric alongside the standard set, and to explain in one line why that metric moves its invoice.

5. Starting without a tag design or a defined intensity denominator. Technically the hardest failure to recover from. Begin metering without settling the naming convention for point identifiers (tags), the equipment hierarchy and the denominator for intensity — units produced, running hours or input weight — and a year later, when someone asks to compare across sites or by product, the data has to be rebuilt, and history cannot be rebuilt retrospectively. If a multi-site roll-out is anywhere on the horizon, fix the convention at the first site and deploy it as the standard.

All five concern something that should have been decided before hardware was purchased. Projects rarely fail because of meter performance; they fail because equipment was bought before the decisions were made.

How to run the project: five steps and how to think about duration

Stating durations in days is largely meaningless, because the schedule is governed not by effort but by two external calendars: shutdown opportunities and the tariff cycle.

Step 1, diagnosis, must happen before any hardware is bought. Lay out twelve months of invoices, split them into energy charge, demand charge, Ft and VAT, and separate seasonal from operational variation. This alone tells you whether you are a plant with demand headroom or a high on-peak share.

Step 2, metering design, exists to prevent failure pattern 5. Its deliverable is a document; equipment selection comes afterwards, and settling the interface with the production management system here removes a great deal of rework.

Step 3, the proof of concept, is restricted to one feeder. The purpose is not technical validation but validation that real improvement actions emerge from this data. If none emerges, the problem lies in metering granularity, analytical framing or the operating model — and a PoC lets you fix it before the roll-out.

Step 4, full deployment, is rate-limited by the annual maintenance shutdown. Software work proceeds in parallel, but CT installation waits for the outage, so the most efficient plan works backwards from the next shutdown and completes steps 1 to 3 before it.

Step 5, embedding, needs a full turn of the seasons. Cooling and HVAC loads differ between wet and dry seasons, and because the Ft is revised every four months, only after three revisions in a year can you speak confidently about underlying performance net of the Ft effect. Conclusions drawn from one or two months risk mistaking a seasonal effect for an improvement.

Ideally step 1 already quantifies the available headroom. If you can say internally that “our on-peak share is 60% today, and shifting 10% of it is worth this much per year”, the capital request reads completely differently; without that estimate it stops dead at “so how much will it save?”. If you outsource the diagnosis, confirm what the deliverable will contain before placing the order.

Summary

First, the bill is not consumption times a unit rate; it is determined by time-of-use pricing, maximum demand within the month, and an Ft you cannot control — and total-volume reduction attacks only half of that structure. Second, metering the incomer is not enough: only at feeder and area level, and then at the individual assets that matter most, does data convert into action. Third, cost has five layers — meters, communications, software, electrical installation and operating effort — and the last is the largest hidden cost precisely because it never appears in a quotation. Fourth, measured electricity data is also the primary data for emissions reporting, so given the trajectory of CBAM and group reporting it carries value a cost calculation alone will not capture. Fifth, failures are almost without exception caused by not deciding, before buying equipment, the things that had to be decided first.

The decision framework is short. Can you explain your on-peak share and your demand profile? Can you define the denominator of your intensity metric? Can you name the person who will look at the data every month? With answers to those three, the investment turns into results. Without them, what you will own is a very good-looking dashboard.

TOMAS TECH is a factory IT/OT integrator based in Bangkok serving Japanese and other international manufacturers across the region, and we supply production management and energy management systems through our own product, PEGASUS. Before any conversation about deploying a system, we are happy to start much earlier — going through your invoices together, or working out where the metering points ought to sit. A metering point design on its own, or a survey of what can be extracted from your existing equipment, is a perfectly good place to begin. We will set out realistic options grounded in the local tariff structure and regulatory framework, so please get in touch through our contact form.

Frequently asked questions

Q1. What exactly does the term energy monitoring system cover?

It spans four layers — measurement, visualisation, analysis and control — and how far up that stack a product reaches varies. Measurement and visualisation alone are usually sold as a power monitoring system; add analysis and it is typically called an EMS or energy management system; add control to prevent contracted demand being exceeded and it becomes demand monitoring or demand control. When comparing quotations, establish which of the four layers each supplier has included before you compare prices.

Q2. What is the first thing to do to reduce a factory’s electricity cost?

Decompose twelve months of invoices into energy charge, demand charge, Ft (the fuel adjustment charge) and VAT, and establish what proportion each represents. If your on-peak share is high, peak shifting will pay; if the demand charge is a heavy share, peak shaving will pay. Beginning with equipment upgrades before you have that read tends to produce weak returns. Rates and the Ft are revised periodically, so verification against the latest ERC publication and your own invoice is essential.

Q3. Does demand monitoring really deliver savings?

The demand charge is set by maximum demand within the month, generally the highest fifteen-minute average. (The settlement interval can differ by contract type and distribution utility; check your supply contract or invoice.) A plant running 24 hours a day for 30 days has 2,880 fifteen-minute intervals, and a single one determines the charge — about 0.03% of the time. However much you conserve through the rest of the month, the demand charge will not fall unless you suppress that interval. Read the other way round, this means that avoiding the simultaneous start-up of the assets that create your peak can produce a result on its own. The size of the effect depends on how large a share the demand charge is on your invoice, so start with the itemised breakdown.

Q4. Can this be retrofitted to existing equipment?

In most cases, yes. Options include retrofitting split-core CTs, taking pulse outputs from existing energy meters, polling existing digital power meters over Modbus/RS-485, extracting data from existing PLCs, and temporary measurement with portable loggers. The bottleneck is not technology but shutdown opportunity: fitting new CTs normally requires the circuit to be de-energised. In a 24-hour plant, aligning installation with the annual maintenance shutdown sets the schedule for the whole project, so complete everything that needs no outage first.

Q5. Do we need ISO 50001 certification?

It is not mandatory. The three tests are whether a customer or the parent company requires it, whether you need a common management framework across several sites, and whether you can carry the effort of achieving and maintaining it. If none applies, adopt the processes the standard requires — baseline setting, defined performance indicators, objectives with action plans, continual review — as internal rules, and certify when the need arises. For reference, plants implementing ISO 50001 have been reported to achieve 12–18% reduction in the first year and around 4% a year sustained for more than a decade, and time to a certifiable state is generally put at 9–18 months; all are global reported figures. The finding of roughly 26% lower energy intensity comes from a study of Chinese manufacturing and is not a figure for Thailand or Vietnam.

References

• Thai electricity tariff trends (Wise Digital): https://www.wisebk.com/asean_news/367315/
• Explanation of the Thai electricity tariff structure: https://capsolar.co.th/en/knowledge/thailand-electricity-tariff
• Explanation of the Ft (fuel adjustment charge): https://capsolar.co.th/en/knowledge/ft-fuel-adjustment-charge-thailand
• BOI Smart and Sustainable Industry measures: https://www.boi.go.th/upload/content/Smart_and_Sustainable_Industry.pdf
• Reference material on Thailand’s energy conservation framework (NEDO): https://www.nedo.go.jp/content/100961522.pdf
• Reporting on Vietnam’s two-part tariff pilot: https://vietnamnews.vn/economy/1725129/ministry-proposes-two-part-electricity-tariff-to-be-piloted-from-2026.html
• Reference information on manufacturing electricity tariffs: https://arcusenergyasia.com/resources/tariffs/manufacturing
• CBAM definitive period and scope: https://carboneer.earth/en/2025/12/cbam-definitive-period-scope-extension/
• Analysis of CBAM’s impact on Vietnam: https://terawatttimes.org/cbam-country-intelligence-vietnam-2026/
• Research on ISO 50001 and energy intensity: https://www.sciencedirect.com/science/article/abs/pii/S0313592625001237
• Research on IoT-based energy management: https://link.springer.com/article/10.1007/s43926-025-00110-y