Factory Power Monitoring System — Cut Thai Electricity Cost

“We installed a factory power monitoring system and we can see a graph of the incoming supply. So how much did the electricity bill actually go down?” Few plant staff in Thailand can answer that on the spot. There is a long distance between meters running on a dashboard and the number on the invoice moving. This article is not an explainer on what power monitoring is. It is about design: how many measurement points to install, where, and which report each number is tied to. We work through Thai tariff structure, a three-tier measurement design, how demand monitoring is actually operated, specific energy consumption, and the variables that move project cost.

Why an energy monitoring system on its own does not reduce your electricity bill

Visibility is not a countermeasure; it is the material for choosing one

The pattern repeats constantly: “let’s start by making it visible,” one meter goes on the incoming supply, a cloud dashboard subscription is signed, and that is the end of it. Six months later what remains is a screen with a row of monthly bars. Nobody can explain why the bill rose in the months it rose, or what worked in the months it fell. Eventually nobody opens it.

The reason is simple. The kWh at the incoming supply is the sum of everything happening in the factory. It rises with output, with ambient temperature, and with a compressor drain trap stuck open. Looking only at the total, you cannot separate causes — and without that you cannot decide what to act on.

Electricity data moves money only when three things are true at once:

1. You have identified which component of the invoice it affects (energy charge, demand charge, or power factor).
2. The equipment or feeder driving that component has been separated out by measurement.
3. The numbers are tied to production records and work procedures, down to who does what.

Miss one and the system runs while the money stays still. Put the other way round: place these three at the entrance to your design and the number of measurement points, the communication method, and the screen layout all follow.

The specific misalignments that show up on Thai sites

Local conditions layer on top. The report going to head office is written in kWh and CO2, while the local administration team reads an invoice denominated in baht. Under TOU (time-of-use) tariffs the same 1 kWh carries a different rate on-peak and off-peak, so it is entirely normal for kWh to fall without the amount falling — or the reverse.

The invoice itself is in Thai, and the format differs between PEA (Provincial Electricity Authority) and MEA (Metropolitan Electricity Authority). Expatriate managers rarely read it, and often accounting simply processes the payment. The demand charge line and the power factor penalty line are printed right there, unread.

And the people who maintain the system are Thai staff. Any mechanism built around Japanese-language screens and manuals stops the day the expatriate’s assignment ends. The same trap repeats in factory IoT and machine monitoring implementation.

A Thai factory’s electricity cost is made of three components

Start by breaking the invoice into three

Before you talk about reducing electricity cost, break down how your invoice is constructed. Under an industrial tariff the amount is set by three components plus a fixed service charge — and the three call for completely different countermeasures.

Only after this breakdown can you decide whether you are “90% energy charge, demand not significant,” or “demand is a high proportion, so levelling comes first,” or “a power factor penalty every month — the capacitors may be dead.” An energy-saving programme that picks countermeasures without this breakdown will almost certainly miss.

Energy charge: understand the TOU time structure

Thai tariffs are built on an average rate approved by the ERC (Energy Regulatory Commission) with the Ft (fuel adjustment charge) added on top. The ERC-approved figure for the May–August 2026 period is an average of 3.95 baht/kWh (excluding VAT), which sits on a base rate of roughly 3.78 baht with an Ft of 0.1623 baht (16.23 satang) on top (the base rate is a rounded figure, so adding the two does not land exactly on the average). The preceding January–April 2026 period was 3.88 baht/kWh, so this period rose. The Ft is revised every four months, three times a year.

That average is across all customers, not your unit price. Industrial rates vary with tariff type and supply voltage. The main industrial categories are Type 3 (30–999 kW) and Type 4 (1,000 kW and above), each offering a choice of flat rate or TOU.

Note: reference values shown as base rates (excluding Ft and VAT). The off-peak band includes designated public holidays.

What deserves attention is that the on-peak/off-peak gap is more than 1.6 times — roughly 1.69 baht per kWh below 22 kV. That makes a second countermeasure viable alongside “use less”: “use the same amount at a cheaper hour.” Batch processes, cleaning stages, compressor tank charging, pre-cooling of temperature-controlled chambers — how much load do you have where timing is flexible? Answering that requires time-band consumption per machine. A single point at the incoming supply will never tell you.

These figures change with voltage class, tariff type, and whether you are supplied by PEA or MEA. Always confirm your own tariff type (Type 3 or Type 4, flat or TOU) and supply voltage against your invoice and supply contract. Do not transplant an article’s numbers onto your own site.

Demand charge: fifteen minutes, once a month, decides the amount

The demand charge is routinely misunderstood even by experienced plant people. What is charged is the highest 15-minute average demand (kW) recorded in the month. Not an instantaneous value — a 15-minute average — and recording that peak once means the whole month is charged at it.

Which window the peak is taken from depends on the tariff. Under TOU, taking the maximum within the on-peak band is generally described as standard, but confirm the band definitions against your contract and invoice. Get this wrong and the alarm threshold design described later is misaligned from the start.

The rates quoted are 210.00 baht/kW-month for Type 3 TOU below 22 kV and 132.93 baht/kW-month for 22–33 kV; higher supply voltage carries a lower rate.

Do the multiplication. Below 22 kV, suppose one month your peak landed 100 kW higher than usual: that alone is over twenty thousand baht for the month, and annualised it is not a figure you wave away. It is a very common story that those 100 kW came from the fifteen minutes during morning start-up when three large machines happened to start together. Output did not rise by a single machine’s worth. Only the bill did.

Power factor penalty: the line item most factories have not noticed

Power factor is the ratio of real power to apparent power. Induction motors, welders and transformer magnetising currents generate reactive power and pull it down. Under Thai industrial tariffs, the portion where power factor falls below approximately 0.85 — where reactive power (kVAR) exceeds 61.97% of real power — is described as charged at 56.07 baht/kVAR, alongside a fixed service charge of 312.24 baht/month.

What makes it awkward is that it occurs while the equipment looks perfectly normal. Typical causes:

• A blown fuse on the capacitor bank, or capacitors that have lost capacitance through ageing
• A failed automatic power factor correction (APFC) controller, or step control misbehaving
• Equipment added without any review of capacitor capacity
• Periods of reduced output where motors idle under light load and power factor deteriorates

In every case the shop floor logs “no abnormality.” You cannot notice it without reading the invoice. Which also means the penalty is the kind of expenditure that can be resolved comparatively quickly once you measure and find the cause — exactly why recording power factor and kVAR belongs in the earliest stage of any power consumption measurement programme.

Setting priorities after the three-way breakdown

First, check whether a power factor penalty is being incurred; if so, the cause is likely a fault and the cost-effectiveness of fixing it is clear. Second, look at the demand charge’s share of the bill; if it is high, identifying when peaks occur and which equipment causes them is the priority. Reducing the energy charge is slower than either — it accumulates from equipment renewal, operational improvement and load shifting. Precisely for that reason you need a mechanism that sustains the effort: specific energy consumption management.

How to design the measurement points of a power monitoring system

Not one point at the supply, not every machine — cut it into three tiers

Measurement point design decides whether an energy monitoring system succeeds, and the classic failure is swinging to an extreme.

Extreme one: the incoming supply only. Minimal cost, but no cause identification. You stay stuck at “it went up this month,” and the activity dies.

Extreme two: meter everything. You learn everything, but meter count, installation cost, shutdown windows, cabling and data volume all shoot up together — and only a subset of the points is ever looked at. Taking shutdowns across the whole plant to install meters for data nobody uses is not a rational investment.

What works is a three-tier design that expands in stages.

Tier 1: first, get to a state where you can reconcile against the invoice

The first job at Tier 1 is to check that your measured values reconcile with the utility’s invoice. A discrepancy here makes every subsequent analysis suspect. Check monthly kWh, the split of time-band kWh (does your on-peak/off-peak boundary match the utility’s definition?), and maximum 15-minute demand.

A surprisingly common issue is a demand aggregation window that does not match the utility’s. Fixed windows (blocks of 15 minutes from the top of each hour) versus a rolling average produce different maxima from the same current waveform. Confirm which definition the utility uses and align your settings. A few kW of discrepancy is not a practical problem in itself, but since these values feed alarm thresholds you should know which definition you are running.

Tier 2: cut it along the lines where ownership divides

There is no single correct way to cut Tier 2, but one criterion works in practice: cut along the boundaries where responsibility for improvement divides — per production line (the line leader owns it), per utility (maintenance owns it), per building (administration owns it). A measurement point with no defined audience ends up with no audience.

Separating utility systems in a Thai factory frequently reveals how large the compressor and HVAC share is. Air leaks are the classic waste, and compressors that keep running at night and at weekends with nothing in production are a strong signal to suspect leakage (there may be a legitimate explanation such as pressure holding or supply to another use, so confirm on site first). One Tier 2 point on the compressor system makes weekend consumption visible. With only the total at the incoming supply, that separation is impossible.

Tier 3: not every machine — top consumers first

Tier 3 adds a focused set of high-consumption machines from the systems Tier 2 flagged. As a Pareto matter, most consumption sits in a small number of machines. Start with the top few, confirm the operating routine works there, then expand. Keeping that order produces results while holding down initial investment.

The order of expansion and the basis for deciding

Staged expansion still stalls without a criterion for moving on:

• Tier 1 reconciles against the invoice and you understand the reality of power factor and demand → move to Tier 2
• Tier 2 shows which system drives the number and improvement owners are assigned → move to Tier 3
• Tier 3 yields per-machine intensity and evidence for renewal investment → widen the set of machines

If nobody is even looking at the Tier 1 data, adding measurement points will not fix it. Repair the operating side before expanding the measurement side.

Measurement methods and communication paths — bringing in existing equipment

Measurement method options

Choose the method from the state of the equipment and the accuracy you need. Nothing requires measuring everything to high accuracy; requirements differ by tier.

The general approach to extracting data from existing equipment maps directly onto the procedure in IoT retrofit for legacy equipment. Power is no exception — bringing old machines into a data system is the same question everywhere.

Communication path: wired or wireless

Wired (Ethernet or RS-485) is stable, but cabling cost varies enormously with distance and route: reusable cable trays, runs between buildings, outdoor routing all change the order of magnitude. Wireless (Wi-Fi, LoRaWAN and similar) keeps installation cost down but is exposed to the plant’s RF environment — metal structures, noise from large machines, interference with existing wireless. Factory wireless design is covered in detail in factory wireless LAN and industrial networks.

In practice, taking Tiers 1 and 2 over wired links and extending Tier 3 over wireless is a workable combination. Holding to the principle that anything where real-time behaviour translates directly into money — demand monitoring above all — belongs on the wired side removes most of the hesitation.

How to decide the sampling interval

Work back from the use case.

• Demand monitoring → an alarm must fire inside a 15-minute window, so a cycle of one minute or less is realistic
• Time-band energy analysis → 15 minutes to one hour is enough
• Specific energy consumption management → match the recording unit of production data (lot, shift, day)
• Equipment anomaly detection (changes in current waveform) → some cases require sub-second sampling

Making everything short-cycle inflates bandwidth and storage. Vary the cycle by tier and purpose.

Demand monitoring is not finished until the shed order is decided in advance

An alarm on its own achieves nothing

A demand monitoring device is installed and an alarm sounds when the threshold is crossed. An enormous number of factories stop there. What actually happens on the floor is usually: “It’s beeping. Is there anything we can switch off?” “The line is running, we can’t stop it.” “Let’s watch it.” Then fifteen minutes elapse and the peak is recorded.

Demand monitoring affects money only when what gets switched off, by whom, and in what order is decided in advance. Document that as a demand-suppression procedure, post it on the floor, and drill it — only then is it a mechanism.

How to decide the shed order

Three criteria for selecting what to switch off:

1. It must not affect quality or safety. Furnace temperature holding, plating tank circulation, cleanroom positive pressure — anything where restarting costs time and money is out of scope.
2. The effect must be significant. Shedding one machine of several tens of kW is more reliable than ten machines of a few kW each.
3. It must recover quickly. You only need to ride out a 15-minute window, so equipment taking an hour to come back is unsuitable.

Loads that fit tend to be HVAC (temporary setpoint change or partial shutdown), compressors (stopping an additional unit while tank pressure has headroom), water heaters, some lighting, and equipment in non-production areas. Stopping the production line itself is the last resort.

Write the procedure in stages; the percentages below are a starting point, and actual thresholds should come from your own history. At the early-warning stage (predicted value reaches roughly 80% of the threshold), monitoring is intensified and additional start-up of large equipment is held — what you decide here is who watches and which terminal the alarm appears on. At stage 1 (predicted value around 90%), pre-designated non-production loads are shut down, which presupposes the equipment list, responsible operator and switch location are committed to paper. At stage 2 (predicted value on course to reach the threshold), utility systems are shut down in addition, and you must have decided in advance who judges the impact on quality. If an exceedance is confirmed, record it, analyse the cause and feed it into prevention for the following month. Only when the record format and the meeting that reviews it are also fixed does the procedure become a mechanism.

The critical point is that it operates on a predicted value. Learning you exceeded the threshold after the window closed is too late. From consumption in the first part of the window, predict the average at the end and raise the alarm as that prediction approaches the threshold. That is the actual function of a demand monitoring device.

Where to set the threshold

Set it by working back from contracted demand (or your target demand value). Watch out for setting it too low: the alarm fires so often that the floor ignores it, and the moment an alarm is treated as “that thing again,” the mechanism is dead.

Start from 12 months of demand history and analyse at what time and under what conditions the highest few peaks occurred. Peaks usually cluster into patterns — Monday morning start-up, the return from an extended holiday, April and May afternoons when cooling load is high, or specific large machines starting together. Once the pattern is known you can sometimes prevent it outright by staggering start-up timing, including interlocks prohibiting simultaneous starts. Before setting up an alarm, first consider whether the event can be designed out.

Knock-on effects for contracted demand

Demand is not only a monthly bill question. As you add equipment, contracted demand itself will need reviewing; conversely, if actual peaks run consistently well below contract, a review may reduce fixed cost. Contract changes carry procedures and conditions, including how a subsequent increase is treated, and need confirming with PEA/MEA and a licensed electrical engineer or contractor. This is not territory an article’s generalities can settle. Being able to walk into that discussion holding measured data is itself one of the practical returns of a monitoring system.

Specific energy consumption (kWh per unit produced) is the heart of energy monitoring

Totals do not sustain an improvement programme

The biggest reason energy-saving programmes fizzle out is that the metric stays as total consumption (kWh). Totals move with output. If the team improves things but output rises and the total rises with it, the floor feels it worked hard for nothing; if production falls and the total falls, that gets logged as “energy saved.” In neither case does the number correspond to the activity.

The answer is specific energy consumption (SEC).

Aligning the denominator is in fact the hardest part

What trips up SEC management is, almost without exception, the denominator. Electricity data is collected automatically, but production counts exist only on a handwritten daily report; shift boundaries do not align with electricity timestamps; changeover timing is not recorded; whether rejects are included differs by department. In that state, doing the division yields nothing meaningful.

So SEC management requires:

• Production records captured on the same time axis as the electricity data (automatically, if possible)
• Product type and lot changeover times recorded
• The aggregation scope of the numerator (kWh) and the processes covered by the denominator matching
• The counting definition documented — good units or units started

This is what it means to connect production management and machine monitoring to power monitoring. Where a platform already captures production results and machine status — as PEGASUS, TOMAS TECH’s production management system, does — computing SEC is simply a matter of putting electricity data onto the same time axis. Deploy a factory EMS in isolation and you are left with manual data entry for the denominator, which hollows out within a few years. On capturing machine running and stoppage data, the acquisition design in predictive maintenance systems applies directly.

Separate out standby consumption

Alongside SEC, the item with the largest practical return is separating running consumption from standby consumption. Overlay the run signal (or a current threshold) with the electricity data and you can extract the power consumed when nothing is being produced.

Do this in a Thai factory and the following tend to surface: equipment that does not stop over the lunch break, conveyors left running through shift changeover, dust collectors running at weekends, tanks still held at temperature with no production. These can be cut without degrading equipment performance and with almost no capital outlay.

Reducing standby consumption affects the total, and depending on conditions affects demand as well. Being an operational improvement, the time to result is short. If you need to show results in the first few months, this is the realistic place to aim.

How to set the baseline

To claim an improvement you need a baseline. Year-on-year comparison for the same month is easy to grasp, but output and ambient temperature both differ, so it does not stand up as evidence.

In practice, place output and, where relevant, variables such as ambient temperature and operating days as explanatory variables, estimate expected consumption, and treat the gap against actuals as the effect. This is the thinking behind energy baselines (EnB) and energy performance indicators (EnPI) in ISO 50001 energy management. Even without rigorous statistical treatment, documenting what you hold constant in the comparison pays off in both head office reporting and third-party verification.

External requirements are now a tailwind for getting energy data in order

The internal case is being strengthened from outside

An investment proposal can be hard to approve on projected electricity savings alone, because the effect cannot be asserted. As of 2026, however, factories in Thailand face a growing volume of data required from outside, and that helps the case.

There are broadly five sources. DEDE (พพ., the Department of Alternative Energy Development and Efficiency) requires annual energy management reports from designated factories and buildings, presented by end use and by system — a direct motive for switching manual tallying to automatic aggregation. ISO 50001 requires energy baselines, performance indicators and procedures for monitoring, measurement and analysis, which translate straight into system design requirements. CBAM requires verified annual emissions reporting for covered goods, which cannot be calculated without electricity consumption at product level. Head offices and customers increasingly ask for Scope 1 and 2 emissions broken down by site and product, and a site total no longer supports an allocation rationale. And using tax measures and BOI incentives requires measured data to select what to invest in and to explain the effect.

DEDE (พพ.) annual energy management report

Under พ.ร.บ. การส่งเสริมการอนุรักษ์พลังงาน (the Energy Conservation Promotion Act), designated factories (โรงงานควบคุม) and designated buildings (อาคารควบคุม) must submit an annual energy management report (รายงานการจัดการพลังงาน) to พพ. Categories divide by the combined capacity of electricity meters or transformers: Group 1 is below 3,000 kW combined, Group 2 is 3,000 kW and above.

พพ. has announced that the submission deadline for the report covering B.E. 2568 (2025) is 31 March B.E. 2569 (2026). Whether your site is covered, and the exact deadline and format each year, must be confirmed against พพ. announcements and your registration status. Details of penalties are not covered here, as this article’s research could not confirm them reliably.

What matters in practice is that many factories still assemble this data by hand every year — transcribing meter readings, cross-checking an equipment list, allocating by end use. If that consumes tens of hours a year, the labour alone forms part of the investment case. And once collection is automatic, data that existed solely for the report becomes usable for day-to-day improvement.

ISO 50001: the standard’s requirements become your design requirements

ISO 50001 is the international standard for energy management systems. Whether or not you pursue certification, its cycle — measure, set a baseline, set targets, act, verify, correct — applies directly to power monitoring system design.

The requirement to identify significant energy uses is the same question as where to place Tier 2 and Tier 3 measurement points. Setting energy performance indicators (EnPI) is the same question as choosing the SEC denominator. Even factories not seeking certification find that borrowing this structure organises the design.

CBAM: the stage where product-level data is required

CBAM (the EU Carbon Border Adjustment Mechanism) entered its definitive phase on 1 January 2026. Covered goods are iron and steel, aluminium, cement, fertiliser, electricity and hydrogen. Exporters from Thailand to the EU face annual emissions reporting verified by an EU-accredited body.

On scale, Kasikorn Research estimates that entry into the paid phase will affect roughly 3.8% of Thailand’s exports to the EU, on the order of 28 billion baht, with iron and steel and aluminium most affected.

CBAM connects directly to power monitoring because emissions must be calculated per product. Plant-wide kWh times a CO2 factor gives a site total, but “emissions per tonne of this product” requires carving out the electricity used to make it. Tier 3 per-machine measurement tied to production records is the precondition, and the machinery of SEC management doubles as the foundation for CBAM compliance and CO2 emissions visualization.

Even outside the covered list, requests from EU- and Western-facing customers for supply chain emissions data (Scope 3) are increasing. “We’re not steel, so it doesn’t apply” no longer holds in every situation.

Tax measures and BOI: support on the investment side

Measures encouraging investment are also in motion. As a cabinet-approved measure, capital investment in certified energy-saving equipment is eligible for a 150% deduction (conditional on the equipment being in operation by December 2028). The BOI has also expanded support for energy efficiency and smart manufacturing.

Eligibility varies with equipment type, certification, the situation of your establishment, and your BOI privileges. Confirm whether you qualify with tax and accounting professionals and with the BOI directly. Still, the existence of such schemes is worth having to hand: a proposal built as “this is the foundation for meeting regulatory requirements, and tax measures exist for it” clears internal approval more easily than one built solely on “electricity cost may go down.”

Six layers that move the cost of a power monitoring system

“One lump sum” cannot be compared

Quotations vary widely with scale and configuration. This article gives no specific figures, because pricing depends on plant configuration and no number can be asserted. Instead, here are the variables that move cost, separated into layers. Ask for quotations broken down this way and you can compare vendors.

How to phase a small start

Phase 1 (Tier 1 plus the operating routine): install meters at the incoming supply and reach a state where you can reconcile against the invoice. Record demand and power factor, establish whether a power factor penalty exists and what pattern the demand peaks follow, and decide who looks at what and where the monthly review happens. If the routine does not run here, moving on changes nothing.

Phase 2 (Tier 2 plus the demand-suppression procedure): add points on major feeders and utility systems to produce a breakdown by system. Document the demand-suppression procedure, decide target equipment and responsible operators, and drill it. Separating standby consumption can start here too.

Phase 3 (Tier 3 plus SEC): add points on the highest-consuming machines and tie them to production records to compute SEC. Set a baseline and put it on a monthly evaluation cycle. If DEDE reporting or CBAM compliance requires per-product figures, build the capability at this stage.

Judge at the end of each phase whether proceeding is worth it. The advantage of this division is that stopping partway does not waste what came before.

What bites in scheduling is not the installation work but when a shutdown window is available. Aligning with extended holidays such as New Year, Chinese New Year or Songkran means short work but a long wait to start. Clear first the locations that split-core CTs can cover without a shutdown, and those readable from existing communication ports, and you can be collecting data and building the routine while you wait. Separating work that requires a shutdown from work that does not, at the planning stage, makes the overall duration predictable.

What to hand over when requesting a quotation

Quotation accuracy is determined by the information you provide; an underinformed quotation always swings high. At minimum, prepare:

• Single-line diagram (current revision, verified against the installed reality)
• Supply contract information (tariff type, contracted demand, supply voltage, PEA or MEA)
• The last 12 months of invoices (showing actual demand and power factor)
• Major equipment list (ratings, quantities, locations, existing meters or communication ports)
• Panel photographs (interior of the receiving panel and main power panels, to check free space)
• Periods and time bands where a shutdown is possible (New Year, Chinese New Year, Songkran and other extended holidays)
• Current state of production data (which system holds it, at what granularity, in what format)
• Reporting requirements (head office format, whether DEDE reporting applies, customer emissions requests)

The last two matter most. Decide the meters while these are still vague and you will later hit “we can’t produce it per product” or “it doesn’t fit the head office format,” and have to redo the work.

Failure patterns, and designing operations that last

Four common failures

1. Building a dashboard and stopping there. The screen is impressive, but who looks at it and when has not been decided, and it connects to no improvement action. The fix is to decide the monthly review forum and its owner before the screen. Something concrete comes first — “in the first-week production meeting each month, report SEC by system in five minutes” — and the screens serve that report. Reverse the order and they go unused.

2. It stops when the person in charge transfers. An expatriate builds it and nobody touches it after the assignment ends. The fix is to prepare screens, procedures and alert destinations in Thai or English and launch with a Thai staff member as primary owner. A mechanism dependent on one individual is always shorter-lived than the expatriate rotation cycle.

3. Measuring only the incoming supply and finishing with “let’s try harder.” Causes do not emerge from a total. Either progress to Tier 2 or, if you will not, invest in something other than measurement — equipment renewal, or direct changes to operating rules.

4. The SEC denominator is not aligned. Departments count production differently, changeover times are unrecorded, timestamps do not line up. The fix is to check the state of production data before beginning power monitoring design. If the denominator cannot be captured, that is where the work starts.

Minimum conditions for a mechanism that lasts

• The audience is defined: assign an owner per system and per line, and name them on the screen or report
• The forum is defined: build it into existing meeting bodies (production meeting, maintenance meeting, monthly report)
• It runs in the local language: screens, procedures and alert text in Thai or English, with head office reporting handled by automatic conversion or a separate report
• The numbers connect to actions: decide a handful of patterns for “what we do when SEC deteriorates”
• It keeps up with change: assign an owner and procedure for updating measurement points and screens when equipment is added or lines change
• Effects are recorded: keep each countermeasure together with the before-and-after numbers, as material for the next investment decision

The last is routinely undervalued and is in fact the most valuable. Once a handful of your own results have accumulated, you no longer need external generalities in your next investment proposal.

Frequently asked questions

How much will a power monitoring system reduce our electricity bill?

There is no single reduction rate. The composition of the bill (energy, demand, power factor), current operating maturity, and scope for equipment renewal make the available range completely different site to site. That said, where the reduction potential lies can be assessed reasonably well even before installation: (1) is a power factor penalty being incurred in the last 12 months of invoices, (2) what proportion of the total is the demand charge, (3) are demand peaks concentrated in specific times and conditions, and (4) how much consumption occurs during non-production hours. All four can be checked with invoices and simple measurement. Confirming them together and then sizing the investment produces less waste than a proposal promising a reduction percentage up front.

Is there any point in measuring only the incoming supply?

There is, but keep the objective limited. Four things stand up on a single incoming-supply point: reconciliation against the invoice, demand monitoring and suppression, power factor anomaly detection, and tracking plant-wide totals over time. Those alone can deliver real returns. What you cannot do is cause analysis — which machine increased, which process lost efficiency. So if you start there, state explicitly to stakeholders that cause analysis is not yet possible and run it against the concrete targets of demand and power factor. Say vaguely that you “made it visible” and expectations will diverge from results.

Our existing equipment is old and has no communication capability. What can we do?

You can measure without any communication capability on the equipment. The basic method is retrofitting a CT and power meter on the supply side of the distribution board, requiring no modification to the machine. If an existing meter has a pulse output you can receive that, and if inverters or multimeters support Modbus RTU (RS-485) you can sometimes read from those. Four things to check on site: (1) mounting space inside the panel, (2) where a CT can be fitted and the primary current value, (3) when shutdown work is possible, and (4) whether existing devices have communication ports and documentation. For the approach to old equipment, the procedure in IoT retrofit for legacy equipment is a useful reference.

What is the difference between a demand monitoring device and a power monitoring system?

A demand monitoring device is specialised in predicting demand within a 15-minute window and raising an alarm. It can be deployed standalone and is relatively inexpensive. A power monitoring system (or energy monitoring system / factory EMS) collects data from multiple points and covers time-band analysis, breakdown by system, SEC management and report output. If demand is your only objective, starting with a demand monitoring device is a rational choice. But if you intend to extend into per-system analysis and SEC management later, deciding how to build the collection platform at the outset reduces rework. Deciding the data outputs first — who receives which report — settles this question.

What will happen to electricity prices in Thailand going forward?

Future prices cannot be predicted. As a verifiable fact, the ERC-approved average rate was 3.88 baht/kWh for January–April 2026 and 3.95 baht/kWh for May–August 2026 (both excluding VAT), so this period rose. The Ft is revised every four months, three times a year, and is affected by fuel prices and exchange rates. The realistic framing is to treat tariff movement as outside your control and work on the variables you can control: consumption, peak, and power factor. The actual unit price applying to you varies with tariff type and supply voltage, so confirm it against your invoice and supply contract.

Do we need to produce data twice — once for head office and once for reporting in Thailand?

With the right design you can consolidate. What you need is to hold source data at the smallest unit (per measurement point, per hour) and keep the aggregation switchable afterwards. kWh and CO2 conversion for head office, results by end use and system for the DEDE report, SEC for internal improvement — build a structure that produces different aggregations from the same source data and duplicate management disappears. Store only pre-aggregated values shaped to one format and you rebuild from scratch the moment another format is requested. For CO2 conversion, also record the source and applicable year of the emission factor you use.

Summary

Whether a power monitoring system succeeds is decided by design, not equipment performance. The key points:

• Start by breaking the invoice into three components: energy charge (TOU time bands), demand charge (15-minute maximum demand), power factor penalty. The three call for different countermeasures and carry different priorities
• Understand the Thai tariff structure: the ERC-approved average for May–August 2026 is 3.95 baht/kWh (excluding VAT; a base rate of roughly 3.78 baht with an Ft of 0.1623 baht on top). TOU on-peak is Mon–Fri 09:00–22:00. The on-peak unit rate is more than 1.6 times the off-peak rate, which makes load shifting viable
• Cut measurement points into three tiers: incoming supply → major feeders and systems → large loads. The incoming supply alone yields no causes; metering everything blows up the cost. Cut Tier 2 along the boundaries where improvement ownership divides
• Demand monitoring is not finished until the shed order is decided: alarm on a predicted value, and document and drill which equipment is shed, by whom, in what order. An alarm alone just watches fifteen minutes go by
• SEC (kWh per unit produced) is the heart of it: totals do not sustain a programme. The hard part is not the numerator but the denominator — whether production records sit on the same time axis decides the outcome
• Turn external requirements into a tailwind: the DEDE (พพ.) annual energy management report, ISO 50001, CBAM in its definitive phase since January 2026, and the 150% deduction for certified energy-saving equipment
• Take quotations across six layers: meters, electrical work, communications, server and cloud, visualisation, operations. Broken down by layer they can be compared, and phasing becomes easier
• Verify every number against your own invoice and contract: rates vary with voltage class, tariff type, and PEA versus MEA. Do not transplant an article’s figures onto your own site

There is no value in measuring as such. Value appears when you know which line on the invoice a measured number affects, when the person and the forum that look at it are decided, and when what to do if it deteriorates is decided. Reverse that order and what remains is an impressive screen and an unmoved amount of money.

TOMAS TECH is a systems integrator based in Bangkok, Thailand, working across factory IT and OT/FA for manufacturers operating in the country. From our experience handling production results and machine status data with the PEGASUS production management system, we are particularly at home designing setups that put electricity data on the same time axis as production numbers. Early-stage conversations are welcome — “we only meter the incoming supply, so what should we measure next?” or “we can’t tell which part of the invoice is driving the cost.” If we can see a recent invoice and your single-line diagram, we can start with a view on which component holds the reduction potential. Get in touch via our contact page.

References

• Power bills to rise slightly in May-Aug period | Nation Thailand
• Factory Electricity Cost in Thailand | CAP Solar
• รายงานการจัดการพลังงาน | Department of Alternative Energy Development and Efficiency (พพ.)
• CBAM enters definitive phase from January 2026 | Nation Thailand
• CBAM impact on Thai exports to EU | Nation Thailand
• Thailand’s Renewed BOI Incentives for 2026-2027 | Emerhub
• ISO 50001 Energy management | ISO
• Carbon rules reshape Thai business as CBAM looms | Bangkok Post