“For our substation equipment, we have relied on an annual shutdown inspection and a thermography inspection by an outside contractor. Recently, though, head office, our insurer and customer audits have started asking whether we do condition monitoring.” We are hearing this more and more from maintenance and electrical staff at Japanese-owned factories in Thailand. What matters in a switchgear condition monitoring implementation is not how many sensors you install. It is deciding, within the time between the point when signs of deterioration become measurable and the point of failure, who will notice and who will decide whether to stop the equipment. To get there, you decide five things: (1) an inventory of equipment and its criticality, (2) mapping deterioration modes to monitoring methods, (3) the division of roles between periodic inspection and continuous monitoring, (4) where the data goes and the decision-making structure, and (5) acceptance testing.
First, a disclaimer. Figures in this article that carry no cited source (the number of panels in the model factory, the time from signs to failure for each type of deterioration, the number of sensors and so on) are assumptions of this article, set for explanatory purposes. They are neither industry averages nor survey results. The specifications of wireless temperature sensors are the manufacturer’s published values and have not been verified by a third party. In addition, what we write about the 2026 edition of the US standard NFPA 70B is based on revision deliberation documents published by NFPA, not necessarily the wording of the final published edition. None of the products or vendors mentioned are TOMAS TECH’s own; they all belong to other companies.
Why Switchgear Condition Monitoring Matters Now
NFPA 70B moved from “recommended” to “standard”, and the 2026 edition is now current
For the maintenance of electrical equipment, there is NFPA 70B, issued by NFPA (the National Fire Protection Association), a private US standards organization. According to an explainer by the industry outlet Reliable Media, with the 2023 edition (effective 16 January 2023) NFPA 70B was reissued as a “Standard for Electrical Equipment Maintenance”, replacing the earlier “Recommended Practice for Electrical Equipment Maintenance”. Under NFPA’s document rules, a recommended practice uses “should” and is written in non-mandatory language, while a standard contains mandatory provisions using “shall” and is written in a form suitable for adoption. The same explainer also notes that the 2026 edition has since superseded the 2023 edition and is the current edition, but that the edition that applies to a facility depends on which edition the authority having jurisdiction, contracts or insurers refer to, and is not necessarily the latest one.
A book retailer’s product description lists the main updates in the 2026 edition (216 pages) as the introduction of the P–F curve method for determining maintenance intervals, new requirements setting deadlines so that equipment repairs are not deferred, more detailed inspection and testing tables, and the addition of a new chapter on transfer switches.
According to the 2026 edition revision deliberation documents published by NFPA, the following points relevant to switchgear condition monitoring are shown (all are wording under deliberation and not necessarily the wording of the final edition).
- Infrared thermography interval: in the maintenance interval table, infrared thermography for “all equipment” is 12 months when the equipment condition is 1 or 2, and 6 months when the condition is 3 (poor). These are treated as intervals not to be exceeded, for example when manufacturer recommendations are not provided. An explainer by an infrared training institution describes the same intent for the 2023 edition (at least every 12 months, and at least every 6 months for condition 3).
- Continuous monitoring notifications are built into equipment condition assessment: the criteria for classifying equipment into conditions 1 to 3 include “no unaddressed notifications from a continuous monitoring system” (condition 1), “notifications from a continuous monitoring system since the last assessment” (condition 2), and “active or unaddressed notifications from a continuous monitoring system” (one of the conditions for condition 3).
- Permanently installed continuous temperature monitoring can satisfy the infrared inspection requirements: a revision was shown that would accept permanently installed continuous temperature measurement devices as satisfying the infrared inspection requirements. The committee described this as an alternative technology that produces the same results as infrared thermography. This is a permission (“permitted”), not a requirement to install them. The annex also mentions infrared windows (IR windows), which provide line of sight without removing covers.
- Design options that take maintenance into account: the annex lists, as design options, infrared windows that allow inspection without exposing workers to energized parts, temperature sensors on critical terminations, ultrasonic sensors on medium-voltage equipment, partial discharge monitoring of critical cables and equipment, and permanently installed condition monitoring devices.
- Inspection intervals no longer than half the P–F interval: a revision permitting the P–F curve method to determine maximum maintenance intervals is shown, along with an annex explanation that “to prevent failure, the interval of time-based preventive maintenance needs to be no more than half the P–F interval.”
- OT cybersecurity: the deliberation documents show a revision that adds operational technology (OT) cybersecurity for network-connected equipment as an element to be included in the electrical maintenance program (EMP).
- Online partial discharge: in the testing tables, “online partial discharge” on medium- and high-voltage transformer windings and “partial discharge” on power cables above 1000 V, among others, are positioned as test items. These are items selected according to the equipment and the maintenance program, and this does not mean they are mandatory for all equipment.
March 2025: a UK substation fire showed “measured, yet not prevented”
In March 2025, a fire broke out at the North Hyde substation in west London, UK, and Heathrow Airport was closed. According to the Associated Press, the report by the UK’s National Energy System Operator (NESO) found that the catastrophic failure of one of the transformers was most likely caused by moisture entering a bushing, causing a short circuit that ignited the oil. It was reported that elevated moisture had been found in a 2018 oil sample at the same substation, but the bushing had not been replaced. The regulator Ofgem began an investigation into whether National Grid Electricity Transmission, the transmission company that owns the substation, had breached its licence conditions.
This is a case of a transmission company’s substation, not of a factory’s substation equipment. Even so, according to the reports, the signs were being measured, yet they did not lead to decisions and action, and this lesson applies directly when considering condition monitoring at a factory.
Standards for medium-voltage switchgear and partial discharge also keep being revised
Standards continue to move as well. IEC 62271-201, the standard for solid-insulation switchgear above 1 kV and up to and including 52 kV, was published in its third edition on 22 July 2026, with changes including alignment with IEC 62271-200, the standard for metal-enclosed switchgear (third edition 2021, Amendment 1 in 2024). IEC 60270, the standard for charge-based measurement of partial discharges, was published in its fourth edition on 5 June 2025. IEC TS 62478, the technical specification covering partial discharge measurement by electromagnetic and acoustic methods, has its 2016 edition current, and a second edition is under development. Note that the switchgear standards IEC 62271-200 and 62271-201 are not standards that require condition monitoring itself.
What it means for factories in Thailand
NFPA 70B is a private US standard, not Thai law. It does not automatically apply to factories in Thailand. However, a global head office’s maintenance standards, insurers’ risk surveys or customer audits may bring it up as one of the standards they refer to. At that point, whether you can explain “we do not have condition monitoring, but this is how we decide our inspection intervals”, or end up with “we monitor, but nobody has decided who does what when an alarm goes off”, is likely to make a large difference to the assessment. Switchgear condition monitoring is also an effort to create the material for such explanations.
What Is Switchgear Condition Monitoring: How It Differs from Periodic Inspection
Thinking in terms of the P–F curve
The basic tool for thinking about condition monitoring is the P–F curve. The condition of equipment gradually worsens as it is used. The annex of the NFPA deliberation documents defines potential failure (point P) as “the point at which it can be detected that the equipment has begun to deteriorate”, and functional failure (point F) as “the point at which it has reached the end of its usable life and no longer operates”. The time from point P to point F is the “P–F interval”.
If you catch the signs within the P–F interval, you can plan a shutdown and take action. If you do not, one day the substation equipment suddenly stops, and the whole factory stops. Periodic inspection and continuous monitoring have the same purpose: catching the signs within the P–F interval. What differs is the probability of catching them, and how much time is left between catching them and taking action.
What the “half rule” means
The idea shown in the annex of the deliberation documents, that “time-based inspection intervals should be no more than half the P–F interval”, can be understood like this. If the inspection interval equals the P–F interval, then with bad luck an inspection ends just before the signs appear, and the next inspection falls just before point F. Even if you find something, there is almost no time left to prepare a response. If the inspection interval is no more than half the P–F interval, there are at least two inspection opportunities within the P–F interval, and when the signs are found, at least half the P–F interval remains before point F, which creates a margin.
Three ways to measure: offline, online periodic and online continuous
Taking partial discharge measurement as an example, let us separate three ways of measuring. They are as follows, according to a paper presented at NETA PowerTest 2016 by engineers from the US testing company American Electrical Testing (posted on the EA Technology website).
- Offline testing: a test in which the equipment is stopped and high-voltage AC is applied directly. It is normally performed as a factory test or acceptance test.
- Online periodic measurement: measurement taken periodically while the equipment is running, using handheld acoustic (ultrasonic) or electromagnetic detectors.
- Online continuous monitoring: sensors, data collectors, monitoring units and analyzers that constantly collect, analyze and report data on the condition of the insulation.
The same paper states that continuous monitoring systems are “not fool proof”. External noise from fluorescent lights, the operation of nearby equipment, construction work and the like can be recorded as signals resembling partial discharge, and the data requires expert analysis and interpretation. Installing continuous monitoring does not remove the need for human judgment; rather, deciding the decision-making structure becomes a prerequisite.
As one form of predictive maintenance
Like vibration monitoring of rotating machines, switchgear condition monitoring is one form of predictive maintenance. How to proceed with predictive maintenance as a whole and how to structure a PoC are covered in “Predictive Maintenance Case Studies: 90-Day PoC and RFP Guide“. The difficulties specific to substation equipment are that the scope of the shutdown needed for inspection is wide and has a large impact on production, and that a failure stops the whole factory and carries the risk of fire and arcing.
Mapping Deterioration Modes to Monitoring Methods
Before choosing a method, decide “which deterioration of which equipment you want to see”. Typical mappings are summarized in the table.
| Target | Main deterioration modes | Signs | Main monitoring methods |
|---|---|---|---|
| Cable connections, busbar connections and circuit breaker contacts in MV switchgear | Loosened fastening, contact deterioration | Temperature rise at connections | Infrared thermography (periodic), IR windows, permanently installed temperature sensors (continuous) |
| Insulation inside MV switchgear | Internal defects such as voids, surface contamination or tracking | Partial discharge | TEV, ultrasonic, UHF (periodic or continuous) |
| MV cables and terminations | Deterioration of solid insulation | Partial discharge | HFCT (clamped on the earth conductor) |
| Inside oil-filled transformers | Localized overheating, discharge, deterioration of insulating paper | Gases in oil | Oil analysis (periodic), online DGA (continuous) |
| Transformer operating condition | Overload, cooling failure | Oil temperature, winding temperature | Taking in signals from existing thermometers |
| Power quality at the receiving point | Grid-side events such as voltage sags | Voltage fluctuations | Voltage and power quality monitoring |
The last row is not deterioration of substation equipment but events coming from the grid side. Its purpose differs from condition monitoring, but since the sensors sit in the same substation room, planning them together is efficient. For how to record voltage sags, see “Voltage Monitoring System for Thailand Factory Power Quality“.
The important point of this table is that no single method sees everything. Temperature sensors are strong at detecting overheating of connections, but cannot see insulation deterioration. Partial discharge monitoring sees insulation deterioration, but cannot see loose connections. The inside of a transformer is hard to grasp with external temperature or partial discharge measurement, so you look at the gases in the oil.
Partial Discharge Monitoring Methods: The Differences Between TEV, Ultrasonic, HFCT and UHF
What partial discharge is
Partial discharge is a small discharge that occurs in only part of the insulation. The NETA paper mentioned above divides partial discharge into two types: “internal partial discharge” occurring in voids and the like within solid insulation, and “surface partial discharge” occurring on the surface of insulation. Surface partial discharge is said to be especially likely with insulation defects, contamination and high humidity. Partial discharge does not cause an immediate failure by itself, but it is generally regarded as a sign that insulation deterioration is progressing.
Four methods
| Method | What it captures | Main targets | Characteristics | Points to note |
|---|---|---|---|---|
| TEV (transient earth voltage) | Electromagnetic pulses appearing on the surface of the panel enclosure | Internal partial discharge and surface tracking in metal-enclosed switchgear | Can be measured without removing covers or touching energized parts | Insufficient sensitivity for problems in solid-insulated cables |
| Ultrasonic (acoustic) | Sound emitted by discharges | Surface partial discharge, corona | Useful for locating the source | Affected by surrounding sounds |
| HFCT (high-frequency current) | Pulse currents flowing in the earth conductor | Cables and terminations | Can be installed by clamping onto the earth conductor | Check the installation position and whether a shutdown is needed |
| UHF (ultra high frequency) | Electromagnetic waves of 300 MHz to 3 GHz | GIS, transformers, cables | Resistant to external noise | Cannot be calibrated to apparent charge (pC) |
TEV: according to the NETA paper, part of the electromagnetic pulse produced by partial discharge inside switchgear appears on the inner surface of the enclosure and travels round to the outer surface through the seams. The paper states that this phenomenon was discovered in the 1970s and named TEV. Because TEV can be detected without removing panel covers and without connecting to energized conductors, it is convenient for detecting partial discharge while the equipment is running. It is useful for finding partial discharge inside switchgear and surface tracking on internal components, while its sensitivity is said to be insufficient for finding problems in solid-insulated cable systems.
Ultrasonic: surface partial discharge and corona emit sound. The same paper states that acoustic (ultrasonic) detection is effective for locating the source. Measurements are often taken near door gaps or ventilation openings (the view of this article).
HFCT: partial discharge current pulses can be measured by clamping an HFCT (high-frequency current transformer) onto the equipment’s earth conductor. It is used to monitor cables and terminations.
UHF: according to a 2019 review paper by researchers at the University of New South Wales in Australia, partial discharge pulses rise very quickly and therefore emit electromagnetic waves with components in the UHF band of 300 MHz to 3 GHz. The UHF method was developed in 1988 for gas-insulated switchgear (GIS) and later applied to power transformers (1997) and cables (1998). Its advantages are resistance to external electromagnetic interference and the ability to avoid low-frequency noise within substations. On the other hand, the lack of a suitable calibration method compared with conventional electrical methods is cited as a challenge.
A “sensitivity check” rather than calibration
With the UHF method, calibration to apparent charge (pC) as in IEC 60270 is not possible. A paper by researchers at the University of Stuttgart in Germany explains that, for UHF measurement in GIS, quality is confirmed by the two-step “sensitivity check” recommended by CIGRE. First, in the laboratory, a pulse is determined that produces a spectrum similar to the signal of a reference defect; then, on site, the same pulse is injected and it is confirmed whether the adjacent sensor can detect it. This is a procedure for GIS and does not apply directly to air-insulated switchgear, but the idea of confirming in acceptance testing “whether the sensor can really pick up partial discharge on site” is useful regardless of the method.
Methods usable online
The review paper mentioned above organizes whether each technique can be used for online monitoring. It states that the IEC 60270 method (current pulse measurement using a coupling capacitor) is not suited to online monitoring, while acoustic methods, HFCT, TEV, VHF/UHF antennas and DGA can be used for online monitoring. Keeping in mind that IEC 60270 is a measurement standard mainly used in laboratory and factory testing, not a standard for online monitoring, will help you avoid misreading specifications.
Lessons from a deployment on 33 kV switchgear
The NETA paper mentioned above reports one case in which continuous partial discharge monitoring was installed on two sets of metal-enclosed switchgear operating at 33 kV in the US (14 existing outdoor panels and 30 new indoor panels) (the system was made by EA Technology). There are several points that are useful when thinking about factory substation equipment.
- Configuration: a host PC (hub), data collectors (nodes), sensors (TEV, ultrasonic microphones, temperature and humidity, high-frequency CTs for cables, and external antennas for noise rejection), and an interface to SCADA (alarms output from the hub as contacts).
- Design constraints: the existing switchgear could only be shut down feeder by feeder, not completely. The manufacturer’s warranty must not be voided. Insulation ratings must not be reduced.
- Approach to alarms: alarms were set for TEV, ultrasonic and cable partial discharge respectively, and an “alarm span”, under which an alarm is valid only if it persists for a certain time or longer, suppressed false alarms caused by external factors. The settings were decided based on the manufacturer’s recommendations and through consultation with the manufacturer (those values belong to that equipment and product and are not general guidelines).
- Network: for security reasons, the owner’s IT department did not allow the hub to be connected to the corporate LAN. Remote monitoring was therefore done with a wireless router and remote access software.
- What was found: an inspection triggered by a rise in ultrasonic readings suspected to be influenced by humidity (not reaching the alarm level) found corona from an insulating barrier board that was rubbing against a 33 kV insulated conductor. In another case, smoke appeared in the station service transformer compartment about three months after an ultrasonic alarm, and the cause was tracking inside the fuse tube on the high-voltage side.
What can be read from this case is that watching the trend has meaning even when the alarm level is not reached, and that even when an alarm goes off, time may be spent hesitating over whether to shut down. The latter is precisely the reason to decide in advance who does what after receiving an alarm.
Temperature Monitoring of MV Switchgear: Thermography, IR Windows and Wireless Sensors
Loosened fastening and contact deterioration at connections show up as a rise in temperature. There are broadly three ways to look at the temperature of MV switchgear.
Infrared thermography (periodic)
This is the method of asking an outside contractor to take images periodically with an infrared camera. The NFPA deliberation documents indicate that, before inspection, the necessary covers should be removed to secure line of sight to the equipment being inspected, and that the temperature difference between the area of interest and a reference area should be recorded. Opening the covers of MV switchgear requires safety procedures and work qualifications, and many panels cannot be opened while in operation. As a result, a situation can arise where “we do thermography inspections, but the critical cable connections are not visible”.
IR windows
This is the method of fitting windows that pass infrared radiation in the door or side panels of the switchgear and imaging the inside without removing covers. The annex of the NFPA deliberation documents also mentions that infrared windows can provide line of sight without removing covers or panels. The position and field of view of the window determine which parts can be seen, so decide which connections you want to see before fitting them. The inspection itself remains periodic.
Permanently installed temperature sensors (continuous)
This is the method of permanently installing temperature sensors on connections and monitoring them continuously. Because routing wiring to high-voltage parts is difficult, wireless sensors are used.
As an example, according to the datasheet of IntelliSAW (ALTANOVA Group), the company’s SAW (surface acoustic wave) temperature sensors are intended for continuous monitoring of hot spots such as busbars, circuit breaker arms and contacts, and cables, and are wireless and require no power supply (passive). As the manufacturer’s published values, it states a measurement range of −25 to +155°C, an accuracy of ±2°C (0 to 80°C) and ±4°C (full range), up to 12 channels per piece of equipment, and an expected operating life of 20 years or more. The same datasheet also clearly states that care must be taken not to overlap frequency bands within the same equipment, and that compliance with local radio regulations is the responsibility of the system integrator and installer. These are the specifications of one company’s product, not values for SAW sensors in general.
Within ASEAN, too, there is a research example in a Vietnamese science and technology journal that applies a passive thermal monitoring system using wireless temperature sensors based on the SAW principle to switchgear, disconnectors, cable connections, ring main units and the like.
Which to choose
As mentioned above, the NFPA 2026 edition revision deliberations showed a direction in which permanently installed continuous temperature monitoring can satisfy the infrared inspection requirements (this is a permission, not a requirement to install). Rough guidelines for choosing can be organized as follows.
- Few panels, and shutdowns are easy to arrange → periodic thermography, plus IR windows as needed
- Many panels whose covers cannot be opened during operation → IR windows
- High-criticality panels where overheating of connections is expected to progress within a few months or less → permanently installed temperature sensors
The three are not mutually exclusive and can be combined. For example, you might continue periodic thermography of all panels while fitting permanent sensors only on the incoming and main feeder panels.
Online Transformer Monitoring: DGA and Standards
Seeing the inside through the gases in the oil
When localized overheating, discharge or deterioration of insulating paper occurs inside an oil-filled transformer, the oil decomposes, generating gases that dissolve in the oil. Analyzing these gases is DGA (dissolved gas analysis). There are two approaches: taking oil samples periodically for analysis, and online DGA, in which an analyzer is permanently installed on the transformer.
Standards that serve as the basis for interpretation
Two main references are used to interpret DGA.
- IEEE C57.104-2019: the current (Active) guide on the interpretation of gases generated in mineral oil-immersed transformers (published November 2019). It covers the purpose and application of DGA, quality checks and limitations, interpretation and norms, identification of fault types, and case examples. A revision project was approved on 22 May 2024, and revision work is under way. It is not a mandatory standard.
- IEC 60599:2022: guidance for interpreting dissolved and free gas analysis for mineral oil-filled electrical equipment in service (fourth edition, published 25 May 2022). In addition to transformers, it includes information on reactors, bushings, switchgear and oil-filled cables. It states that the results are only guidance and must be accompanied by appropriate engineering judgment.
The transformer’s own trend matters more than absolute values
According to an explainer by members of the IEEE C57.104 revision working group, the norms in the 2019 edition were created from a dataset of 313,076 transformers and 1,391,436 DGA reports (about 310,000 transformers and about 1.39 million reports), based on data collected from 18 contributors. The norms in the first edition of 1978 were based on results from a single region in the 1970s and are said to have remained almost unchanged for about 40 years. The 2019 edition also added tables of the amount and rate of change between consecutive samples.
What can be read from this is the idea that it is important to look at the transformer’s own trend (how much, and how fast, it has changed). The value of online DGA lies in capturing the speed of change that periodic analysis cannot see. Conversely, there are also transformers that change slowly enough for periodic oil analysis to track the trend sufficiently. In the North Hyde case, too, the signs themselves had been caught in an oil sample. What is needed, as much as measuring, is a mechanism that connects the measurement results to decisions on action.
Taking in oil and winding temperature
Many transformers already have oil temperature gauges and winding temperature gauges. Before adding new sensors, simply taking these existing signals into the monitoring system lets you track trends in overload and cooling failure. For how to collect energy and load data from substation equipment, see “Power Meter Data Collection: An RFP for Trusted 15-Minute Data“. Looking at load data alongside temperature makes it easier to tell whether a temperature rise is due to load or to an equipment abnormality.
Dividing Roles Between Inspection and Continuous Monitoring by P–F Interval (Calculation for Model Factory S, Assumed Model)
From here, we calculate how to divide roles between periodic inspection and continuous monitoring using an assumed model factory. All figures below are explanatory values based on the assumptions of this article.
Assumptions (Model Factory S)
- A Japanese-owned parts factory in central Thailand, with 12 MV switchgear panels for incoming power and 3 transformers (assumption of this article)
- For each deterioration mode, the P–F interval from the point when signs become measurable to failure is set at 24 months, 12 months, 6 months, 3 months and 1 month (assumption of this article)
- We assume that the onset of deterioration is uniformly random, that an inspection is over in an instant, and that if signs are present they are always found. In this case, the probability of catching the signs is “the probability that an inspection falls within the P–F window”, which is the P–F interval ÷ the inspection interval (capped at 100%)
As examples of P–F intervals, this article assumes that the short ones correspond to the progression of overheating due to loose connections or the progression of surface discharge, and the long ones to slow deterioration of transformers. Actual P–F intervals vary greatly depending on the equipment and environment, and the values in this article are assumptions for explanation.
Step 1: Inspection intervals required by the P–F half rule
Following the idea that “inspection intervals should be no more than half the P–F interval”, the required inspection intervals and annual number of inspections are as follows.
| P–F interval (assumed) | Required inspection interval (÷2) | Inspections per year |
|---|---|---|
| 24 months | 12 months | 1 |
| 12 months | 6 months | 2 |
| 6 months | 3 months | 4 |
| 3 months | 1.5 months | 8 |
| 1 month | 0.5 months (about 2 weeks) | 24 |
The calculation is: inspection intervals of 24÷2=12, 12÷2=6, 6÷2=3, 3÷2=1.5 and 1÷2=0.5 months, and annual counts of 12÷12=1, 12÷6=2, 12÷3=4, 12÷1.5=8 and 12÷0.5=24.
Step 2: Probability of catching the signs with inspections every 12 months and every 6 months
Next, let us see how well the signs can be caught at the current inspection frequency.
| P–F interval (assumed) | Every 12 months | Every 6 months |
|---|---|---|
| 24 months | 100% | 100% |
| 12 months | 100% | 100% |
| 6 months | 50% | 100% |
| 3 months | 25% | 50% |
| 1 month | about 8.3% | about 16.7% |
The calculation is: 6÷12=50%, 3÷12=25%, 1÷12=0.0833…, about 8.3%, 3÷6=50%, and 1÷6=0.1666…, about 16.7%. If the P–F interval is equal to or longer than the inspection interval, at least one inspection falls within the window, so the probability is 100% (the cap).
There is a caution on how to read this table. When a 12-month P–F is covered by inspections every 12 months, the probability is 100%, but only one inspection falls within the window. With bad luck, the finding comes just before point F, leaving no margin to plan a shutdown or arrange parts. The “half rule” is satisfied only for deterioration with a P–F of 24 months or more under 12-monthly inspections, and 12 months or more under 6-monthly inspections. The 100% in the table only indicates that the signs “can be found”, and does not guarantee that action will be “in time”.
On the other hand, deterioration with a 3-month P–F is caught only 25% of the time with inspections every 12 months, and only 50% even with inspections every 6 months. Deterioration with a 1-month P–F is caught about 8.3% of the time every 12 months, and about 16.7% even every 6 months.
Step 3: Number of sensors for 12 panels and 3 transformers (assumed configuration)
Assuming continuous monitoring is used to cover deterioration with short P–F intervals, we provisionally set a sensor configuration for Model Factory S.
- Temperature: 3 phases × 1 location = 3 points at the cable connections of each panel, so 12 panels × 3 = 36 points. Adding 3 points per panel at the busbar connections gives 12×3=36 points, for a total of 72 points
- Partial discharge: 2 points per panel, 1 TEV and 1 ultrasonic, so 12×2=24 points. HFCTs for cables only on the 4 circuits of the incoming and main feeders (assumption of this article), so 4 points. The partial discharge total is 24+4=28 points
- Transformers: online DGA only on the single most critical unit (assumption of this article). Oil and winding temperatures for all 3 units are taken in from existing signals (taking in existing signals is not included in the point count below)
- Total: 72 temperature points + 28 partial discharge points = 100 points, plus 1 online DGA unit
Step 4: If you try to satisfy the half rule through inspection
To pick up deterioration with a 3-month P–F through inspection, 8 inspections a year are needed, as shown in Step 1. Increasing inspections involving shutdowns to 8 a year is hardly realistic for substation equipment because of shutdown constraints. This part will be covered either by online periodic measurement that requires no shutdown (handheld TEV and ultrasonic detectors) or by continuous monitoring.
How to read this calculation
Three points can be drawn from this calculation.
- Deterioration with a long P–F may be adequately picked up by current periodic inspections. Slow deterioration of transformers (assumption of this article) may be trackable through periodic oil analysis and inspection. There is no need to put everything under continuous monitoring.
- For deterioration with a P–F of a few months or less, continuous monitoring is more realistic than increasing inspections. This is especially true for substation equipment with shutdown constraints.
- Even if you install 100 sensors, they are meaningless unless it has been decided who does what on receiving an alarm. The flow of decisions and actions needs to be designed before deciding the number of points.
Where the Data Goes and the Decision-Making Structure: SCADA, Historian, IEC 61850 and OT Security
The data flow
Condition monitoring data passes from the sensors through a gateway (or data collector) into SCADA or a historian, and reaches the person in charge as an alarm. What you decide at each point in this flow determines the success or failure of the implementation.
- Sensors and gateway: the types and number of sensors, communication with the gateway, and power for the gateway.
- SCADA: real-time display and alarms. As in the 33 kV case mentioned above, there is also the simple form of just outputting alarms to SCADA as contacts.
- Historian: for storing trends over long periods and tracking changes that do not reach alarm levels. Partial discharge and temperature often only become meaningful when you look at the trend, so the design of the historian is important. How to choose a historian is covered in detail in “Industrial Historian Selection: RFP and FAT/SAT in Thailand“.
- Time: to line up alarms from multiple sensors and the operating records of protection relays to find a cause, the clocks need to agree. For the approach to time synchronization, see “Factory Time Synchronization: PTP vs NTP and Acceptance Tests in Thailand“.
What to output over IEC 61850
In the power sector, the communication standard IEC 61850 is used. IEC TR 61850-90-3:2016 is a technical report on using IEC 61850 for condition monitoring, diagnosis and analysis, and it covers communication with sensor networks and the exchange of information with asset management systems.
According to 2010 presentation material by IEC 61850 expert Karlheinz Schwarz, the supervision logical nodes in IEC 61850-7-4 include SPDC for partial discharge monitoring and diagnosis, STMP for temperature supervision, and SIML for supervision of liquid insulating media (we have not checked whether the names have changed in the latest edition). His August 2026 blog shows an example in which STMP holds data such as temperature values and alarms on the rate of temperature change. If an RFP just says “IEC 61850 compliant”, interpretation will vary depending on who receives it. If you write down which logical nodes (SPDC, STMP, SIML and so on) are used to output which data, you will receive proposals that are easier to compare.
Alarm → who → does what
Most important of all is the flow after an alarm. Decide at least the following three stages.
| Stage | Content | What to decide |
|---|---|---|
| Caution | The alarm level is not reached, but the trend has changed | Who reviews the trend, for example weekly, and how much change escalates to the next stage |
| Alarm | The alarm level has been exceeded for a certain time or longer | The first responder, who to contact and within how many minutes, and how to check on site |
| Emergency | The alarm persists, or alarms appear on multiple sensors at once | Who decides whether to transfer load or shut down, and the procedure for calling in outside experts |
Because alarm levels and the “certain time or longer” duration differ by product and equipment, they are adjusted during the commissioning period based on the manufacturer’s recommendations and expert opinion. For handling false alarms and setting thresholds, the approach in “Equipment Anomaly Detection PoC: 90-Day Acceptance Criteria for Thai Factories” can be applied.
Interpreting partial discharge data requires specialist knowledge. If no one in-house can make the judgment, it is realistic to sign, at the same time as the implementation, a contract under which the data is sent to outside analysis experts and you receive periodic reports.
OT security
As mentioned above, in the NFPA 70B 2026 edition revision deliberations, a revision adding OT cybersecurity to the elements of the electrical maintenance program was passed. In the 33 kV case, too, the IT department did not allow the monitoring hub to be connected to the corporate LAN. If you connect the monitoring system to the factory network, agree in advance with the IT department on network segregation, the method of remote access, the software update procedure and account management. If you find out later that “it cannot be connected”, you will have to redo the remote monitoring design.
Issues Specific to Thailand and ASEAN
Grid outage statistics do not represent the failure risk of on-site equipment
According to the 2024 annual report of Thailand’s Provincial Electricity Authority (PEA) (the AI-translated English version), SAIFI in 2024 (the annual number of interruptions per customer) was 1.17, 0.29 fewer than in 2023, and SAIDI (the annual interruption duration per customer) was 21.03 minutes, 6.55 minutes shorter than in 2023. Working back from the differences, the 2023 values were SAIFI 1.17+0.29=1.46 and SAIDI 21.03+6.55=27.58 minutes (calculated by this article).
However, PEA’s sustainability page notes that its SAIFI and SAIDI exclude the three southernmost provinces, and exclude events resulting from severe accidents, force majeure, catastrophes and severe interruptions from power generation sources (we have not been able to confirm whether the same conditions apply to the figures in the annual report). At least the indicators PEA publishes on its sustainability page are treated as averages that exclude major events, and they do not represent the supply risk of an individual factory.
More importantly, grid outages and failures of on-site substation equipment are different things. However stable the grid may be, if overheating progresses at the cable connections of your own MV switchgear, the factory can stop. The health of on-site equipment is something only you can monitor.
Law: inspection, maintenance and records are required, but the frequency is not written in the text of the ministerial regulation
The Thai Ministry of Labour’s ministerial regulation on occupational safety and health concerning electricity (2015, B.E. 2558) requires employers to inspect and maintain electrical systems and equipment, and to have a registered person or similar prepare and certify records of the inspection results so that safety inspectors can check them. The criteria, methods and conditions are to be set by notification of the Director-General. It also requires keeping an on-site electrical circuit diagram certified by an engineer or similar.
However, the text of this ministerial regulation does not state the inspection frequency. For frequency, you need to check the notifications of the competent authority, as well as insurance and head office standards (the view of this article). In addition, this ministerial regulation does not require online monitoring of partial discharge or temperature. An overview of statutory inspections of electrical equipment in Thailand is given in “Thailand Factory Electrical Equipment Statutory Inspection“. It is natural to position condition monitoring not as a replacement for statutory inspections, but as something that fills the gaps between them.
Environment: heat and humidity, dust, salt damage and small animals
Substation rooms in Thai factories can be exposed to an environment of hot and humid climate, dust, salt damage in coastal areas and intrusion by small animals. The NETA paper mentioned above states that surface partial discharge is especially likely with contamination of insulation and high humidity. In the 33 kV case as well, a rise in ultrasonic readings suspected to be influenced by humidity triggered an inspection. In partial discharge monitoring, recording temperature and humidity at the same time makes it easier to tell whether a change in readings is due to equipment deterioration or to the environment. The state of air conditioning, ventilation and sealing of gaps in the substation room is also worth reviewing together with the monitoring plan.
Frequencies of wireless sensors
Wireless temperature sensors and gateways use radio waves in specific frequency bands. Whether they can be used in Thailand needs to be checked against the regulations of the National Broadcasting and Telecommunications Commission (NBTC). The manufacturer’s datasheet mentioned above also states that compliance with local radio regulations is the responsibility of the system integrator and installer. In the RFP, state as a condition that Thai radio certification has already been obtained.
Operations: 24-hour first response, outside experts and procedures in three languages
On the operational side, it is realistic to decide the following points before implementation (the view of this article).
- 24-hour first response: who receives alarms at night or on holidays. The maintenance duty staff, security guards or an outside monitoring service.
- Outside analysis experts: who interprets the partial discharge data. The scope of the contract, the frequency of reports and the response time in an emergency.
- Procedures in three languages: prepare them in Thai, Japanese and English so that Thai engineers, Japanese managers and outside contractors can read the same procedures.
- Planning of shutdown work: whether shutdowns for installing sensors and taking action can be arranged feeder by feeder. Plan work that needs a full shutdown to coincide with long holidays and the like.
12 Items to Include in a Switchgear Condition Monitoring RFP
These are the minimum items to include for switchgear condition monitoring when requesting quotations from equipment manufacturers or system companies.
- List of target equipment and criticality: for each panel, transformer and cable, what stops if it stops, its rating, year of manufacture and the unit by which it can be shut down
- Deterioration modes and methods to monitor: write the mapping of deterioration modes to methods, such as temperature for overheating of connections, partial discharge (TEV, ultrasonic, HFCT, UHF) for insulation deterioration, and DGA for the inside of transformers
- Approach to dividing roles with inspection: the assumed P–F intervals, the scope picked up by periodic inspection and the scope covered by continuous monitoring
- Sensor specifications and installation: measurement range and accuracy (including the manufacturer’s conditions), installation positions, the scope of shutdown needed for installation, and not compromising the manufacturer’s warranty or insulation ratings
- Radio certification and frequencies: Thai radio certification already obtained, and a design in which frequencies do not overlap within the same equipment
- Noise countermeasures and suppression of false alarms: the method of removing external noise (external antennas and so on), a function that raises an alarm only when the condition persists for a certain time or longer, and simultaneous recording of temperature and humidity
- Data output: contact outputs to SCADA, Modbus or OPC UA, and when outputting over IEC 61850, the logical nodes (SPDC, STMP, SIML and so on) and a list of data
- Historian and time: the retention period and resolution of trends, how timestamps are applied and the time synchronization method
- Alarm and decision-making structure: definitions of the caution, alarm and emergency stages, notification recipients and methods, and the procedure for initial setting and adjustment of alarm levels
- Analysis and reporting: the scope of analysis by outside experts, the frequency and format of periodic reports, and the response time in an emergency
- OT security: network segregation, the method of remote access, account management and the software update procedure (subject to approval by your own IT department)
- FAT/SAT and maintenance: the test items in the next section and how to set pass criteria, a maintenance contact in Thailand, response times for failures, and documents and training in Thai, Japanese and English
Items 1, 6 and 9 are especially important. Without 1, the number and placement of sensors in proposals cannot be compared. Without 6, false alarms may continue after operation begins, and there is a risk that nobody will look at the alarms anymore. Without 9, the backbone of this article, “who notices and who decides whether to stop the equipment”, stays undecided and only sensors get installed.
What to Check in Condition Monitoring FAT/SAT
In the factory acceptance test before shipment (FAT) and the on-site test after installation (SAT), what you confirm is not the catalog specifications, but “whether the sensors can really pick up signs in your own substation room”, “whether alarms reach the right people” and “whether false alarms are within an acceptable range”.
| Test item | Method | How to set the pass criteria | Main stage |
|---|---|---|---|
| Mapping of sensors to channels | Change the signal of each sensor one at a time and confirm that the screen and historian tags correspond to the correct panel and phase | No mix-ups at any point | FAT and SAT |
| Temperature sensor readings | Compare with a reference thermometer. On site, compare operating values with another method such as IR windows | Within the manufacturer’s published accuracy | FAT and SAT |
| Sensitivity check of partial discharge sensors | Inject a simulated signal such as a calibration pulse and confirm that the sensor and measurement system can detect it. For UHF, do this as a sensitivity check rather than calibration | The defined simulated signal can be detected | FAT and SAT |
| Background noise | Measure for a certain period in the operating substation room and record the noise level | A record is kept and serves as the basis for initial alarm levels | SAT |
| False alarm test in a noisy environment | Operate nearby lighting, inverters, radios and the like, and confirm that no false alarms occur | False alarms are within the acceptable range, for example through judgment based on persistence for a certain time or longer | SAT |
| Alarm delivery | Raise a simulated alarm and confirm that it reaches the designated people via SCADA, email, phone and so on | Delivered within the defined time and a record is kept | FAT and SAT |
| Data output | Check contacts to SCADA, Modbus, OPC UA and IEC 61850 logical node data | Matches the data list in the RFP | FAT and SAT |
| Time | Compare timestamps of alarms and events with a reference time | Within the defined error | SAT |
| Communication loss and recovery | Cut the power or communication of the gateway and check the alarm and how data gaps after recovery are handled | The loss raises an alarm and gap periods can be identified | FAT and SAT |
| Radio interference | Confirm that sensors on adjacent panels do not interfere with each other | Stable readings at all points | SAT |
| Confirmation of shutdown scope | Confirm that the scope of shutdown needed for installation and maintenance matches the prior plan | No unplanned shutdowns | SAT |
| Drill of the decision procedure | Run through, as a tabletop or live exercise, from a simulated alarm to first response, on-site checking and the decision to stop | Reach a decision within the defined time, following the procedure | SAT |
Because pass criteria values differ by equipment and product, set them at the RFP stage. It is likely realistic to adjust partial discharge alarm levels after the background measurement in SAT and the commissioning period after operation begins. The last item, “drill of the decision procedure”, may be unusual as a test item, but it is the very backbone of this article. Running through it once at the point of acceptance is likely to be of great value.
A 90-Day Plan for Switchgear Condition Monitoring Implementation
Days 0 to 30: Inventory of equipment and assigning criticality
- List the substation equipment (MV switchgear, transformers, cables) and organize what stops if each one stops
- Collect past inspection records, thermography inspection reports and failure history
- Confirm the unit by which shutdowns can be made (feeder by feeder, or full shutdown) and the next shutdown opportunity
- Confirm the standards required by head office, insurers and customers
Days 31 to 60: Designing deterioration modes, methods and the division of roles with inspection
- Starting with high-criticality equipment, decide the mapping of deterioration modes to monitoring methods
- Replace this article’s assumed model with your own equipment and, from the assumed P–F intervals, decide the scope picked up by periodic inspection and the scope covered by continuous monitoring
- Try online periodic measurement of the main panels with handheld TEV and ultrasonic detectors, and grasp the current state and noise levels
- Decide where the data goes (SCADA, historian) and the network policy with the IT department
Days 61 to 90: Decision-making structure and RFP
- Decide the caution, alarm and emergency stages, and the roles of the first responder, the decision-maker and outside experts
- Write the 12 RFP items, and decide the FAT/SAT items and how to set the pass criteria
- Draft the outline of the procedures in Thai, Japanese and English
The deliverables you want to have in hand at the end of the 90 days are six: (1) the equipment list and criticality, (2) the mapping table of deterioration modes to methods, (3) the approach to dividing roles between inspection and continuous monitoring, (4) the data flow and network policy, (5) the alarm and decision-making structure, and (6) the RFP and FAT/SAT criteria.
Frequently Asked Questions
What is switchgear condition monitoring, and how does it differ from periodic inspection?
Switchgear condition monitoring is an effort to measure signs of deterioration (temperature, partial discharge, gases in oil and so on) in MV switchgear, transformers, cables and the like, and to take action before failure. Whereas periodic inspection checks the condition at fixed intervals, condition monitoring includes periodic measurement during operation and continuous monitoring. Deterioration whose P–F interval (from the point when signs become measurable to failure) is short compared with the inspection interval is easily missed by periodic inspection. In this article’s assumed model, the probability of catching deterioration with a 3-month P–F with inspections every 12 months was 25%.
How do I choose a partial discharge monitoring method (TEV, ultrasonic, HFCT, UHF)?
Choose according to the equipment and deterioration you want to see. TEV is used for internal partial discharge in metal-enclosed MV switchgear, ultrasonic for surface discharge, corona and locating the source, HFCT clamped onto the earth conductor for cables and terminations, and UHF for GIS and the like. TEV is said to have insufficient sensitivity for problems in solid-insulated cables, and UHF cannot be calibrated to apparent charge, so an on-site sensitivity check is needed. No single method sees everything, so combining them is common.
For temperature monitoring of MV switchgear, are wireless sensors or thermography better?
Think in terms of a combination rather than one or the other. As a rough guideline: periodic thermography if there are few panels and shutdowns are easy to arrange, IR windows if many panels cannot have their covers opened during operation, and permanently installed temperature sensors for high-criticality panels where overheating is expected to progress in a short period. When using wireless sensors in Thailand, the NBTC regulations need to be checked.
Is online transformer monitoring (DGA) necessary?
It depends on the criticality of the transformer and how its deterioration progresses. Some transformers change slowly enough that periodic oil analysis can track the trend. On the other hand, for transformers whose stoppage stops the whole factory, or where you want to capture the speed of change, online DGA is a candidate. IEEE C57.104-2019 and IEC 60599:2022 are used for interpretation, and it is important to look at the transformer’s own trend rather than absolute values. One approach is to start by taking in signals from existing oil temperature and winding temperature gauges.
Is NFPA 70B relevant to factories in Thailand?
NFPA 70B is a private US standard, not Thai law, and it does not automatically apply to factories in Thailand. However, it can become one of the standards referred to by a global head office’s maintenance standards, insurers or customer audits. According to the 2026 edition revision deliberation documents, the P–F curve method, the ability of permanently installed continuous temperature monitoring to satisfy infrared inspection requirements, OT cybersecurity and other points are shown (not necessarily the wording of the final edition). Statutory inspections under Thai law should be checked against the Ministry of Labour’s ministerial regulation and related notifications.
What does switchgear condition monitoring cost to implement?
The cost varies greatly depending on the number of target pieces of equipment, the types and number of sensors, the shutdowns and work needed for installation, the gateway and communication, integration with SCADA and historians, contracts for analysis by outside experts, and the scope of maintenance. In this article’s assumed model, we set a configuration of 72 temperature points and 28 partial discharge points, 100 points in total, plus 1 online DGA unit for 12 MV switchgear panels and 3 transformers, but there is no need to fit every panel at the same density. Narrowing down the scope that needs continuous monitoring based on criticality and P–F intervals is the starting point for thinking about cost.
Summary
- The value of switchgear condition monitoring lies not in the number of sensors, but in deciding, within the time between signs of deterioration and failure, who will notice and who will decide whether to stop the equipment.
- The NFPA 70B 2026 edition revision deliberation documents show the idea that “inspection intervals should be no more than half the P–F interval” and that permanently installed continuous temperature monitoring can satisfy the infrared inspection requirements. It is a private US standard and not Thai law, but head office, insurers and customer audits may refer to it.
- Map deterioration modes to methods: temperature for overheating of connections, partial discharge (TEV, ultrasonic, HFCT, UHF) for insulation deterioration, and DGA for the inside of transformers. No single method sees everything.
- In this article’s assumed model, the probability of catching deterioration with a 3-month P–F is 25% with inspections every 12 months and 50% even every 6 months. Satisfying the half rule through inspection for deterioration with a 3-month P–F requires 8 inspections a year, so because of shutdown constraints it is realistic to cover it with continuous monitoring or online periodic measurement.
- Look at data trends in a historian, and decide the flow after an alarm (caution, alarm, emergency) and who makes the decision. If you output over IEC 61850, specify down to the logical nodes, and agree on OT security with the IT department in advance.
- In Thailand, grid outage statistics do not represent the failure risk of on-site equipment, and the Ministry of Labour’s ministerial regulation does not state the inspection frequency in its text. Wireless frequencies need to be checked against NBTC regulations.
- At acceptance, confirm sensor mapping, sensitivity checks, false alarms in a noisy environment, alarm delivery and a drill of the decision procedure in your own substation room.
TOMAS TECH supports Japanese-owned factories in Thailand from the side of integrating switchgear condition monitoring data into SCADA and historians and designing trend dashboards and alarm flows. You can consult us from the consideration stage, such as inventorying substation equipment, sorting out which method to fit on which panel, examining where the data goes and the alarm structure, preparing the RFP, and attending FAT/SAT. Even if you are “not yet at the stage of choosing equipment, but want to sort out how to think about monitoring our substation equipment”, please feel free to contact us via our contact form.
References
- Reliable Media “NFPA 70B Compliance”: https://reliamag.com/guides/nfpa-70b-compliance/
- NFPA “NFPA 70B (2026 edition) First Revision document”: https://docinfofiles.nfpa.org/files/AboutTheCodes/70B/70B_F2025_EEM_AAA_FD_FRStatements.pdf
- NFPA “NFPA 70B Public Input and committee responses”: https://docinfofiles.nfpa.org/files/AboutTheCodes/70B/70B_F2025_EEM_AAA_FD_PIResponses.pdf
- NFPA “NFPA 70B Second Draft ballot results”: https://www.nfpa.org/api/files?path=%2Ffiles%2FAboutTheCodes%2F70B%2F70B_F2025_EEM_AAA_SD_BallotFinal.pdf
- Fire-Police-EMS “NFPA 70B 2026 edition”: https://www.fire-police-ems.com/NFPA70B-2026.shtml
- Infrared Training Center “NFPA 70B 2023: New Guidelines for Electric Inspections”: https://www.infraredtraining.com/en-US/home/resources/blog/nfpa-70b-2023-new-guidelines-for-electric-inspections/
- Grace Technologies “NFPA 70B 2026 Updates” (manufacturer’s description): https://www.graceport.com/blog/nfpa-70b-2026-updates-new-requirements-for-thermal-monitoring-maintenance-programs-and-more
- Associated Press (via WSLS) “Fire that shut Heathrow was caused by a preventable technical fault known for years, report finds”: https://www.wsls.com/business/2025/07/02/fire-that-shut-heathrow-was-caused-by-a-preventable-technical-fault-known-for-years-report-finds/
- Estonian Centre for Standardisation “IEC 62271-201:2026”: https://www.evs.ee/et/iec-62271-201-2026
- IEC “IEC 62271-200:2021”: https://webstore.iec.ch/publication/63466
- IEC “IEC 62271-200:2021+AMD1:2024 CSV”: https://webstore.iec.ch/en/publication/96236
- IEC “IEC 60270:2025”: https://webstore.iec.ch/en/publication/65087
- IEC “IEC TS 62478:2016”: https://webstore.iec.ch/publication/25740
- genorma “IEC TS 62478 ED2”: https://genorma.com/en/standards/iec-ts-62478-ed2
- Chai et al. “Application of UHF Sensors in Power System Equipment for Partial Discharge Detection: A Review” (Sensors 2019): https://pmc.ncbi.nlm.nih.gov/articles/PMC6427733/
- Hoek, Coenen et al. “Fundamental Differences of IEC and UHF Partial Discharge Measurement” (CMD 2008): https://www.ieh.uni-stuttgart.de/dokumente/publikationen/2008_Coenen_CMD_Fundamental-Differences-of-IEC-and-UHF.pdf
- Rockwell, Horowitz and Smith “Application of On-Line PD Technology for MV Switchgear” (NETA PowerTest 2016): https://eatechnology.com/media/zxdmcm1w/neta-powertest-2016-paper_application-of-on-line-pd-technology-for-mv-switchgear.pdf
- IntelliSAW “SAW Temperature Sensor” datasheet (manufacturer’s published values): https://doble.com/wp-content/uploads/LP-temperature-sensor.pdf
- Vietnam Journal of Science and Technology “Study on SAW wireless passive temperature monitoring” (Vol.65 No.10B): https://vjol.info.vn/index.php/most/article/view/85363
- IEEE “IEEE C57.104-2019”: https://standards.ieee.org/ieee/C57.104/7476/
- IEEE “PC57.104 (revision project)”: https://standards.ieee.org/ieee/C57.104/11641/
- Transformer Technology “The Data Behind the Numbers: IEEE C57.104-2019”: https://transformer-technology.com/article-hub/the-data-behind-the-numbers-ieee-c57-104tm-2019-dga-interpretation-guide/
- IEC “IEC 60599:2022”: https://webstore.iec.ch/publication/66491
- Estonian Centre for Standardisation “IEC TR 61850-90-3:2016”: https://www.evs.ee/en/iec-tr-61850-90-3-2016
- Karlheinz Schwarz “IEC 61850 presentation material” (March 2010): https://content.nettedautomation.com/n/download/pub/DT-Tampa-Pres_2010-03-24.pdf
- Karlheinz Schwarz “How IEC 61850 information models are…” (August 2026): https://blog.nettedautomation.com/2026/08/how-iec-61850-information-models-are.html
- PEA “Annual Report 2024” (AI-translated English version): https://www.pea.co.th/sites/default/files/annual-report/2025/PEA%20AR%202024_ai%20translation.pdf
- PEA “Security and Stability”: https://www.pea.co.th/en/sustainability/social/security-stability
- Thai Ministry of Labour “Ministerial Regulation on Occupational Safety and Health Concerning Electricity (B.E. 2558)” (copy posted by Naresuan University, in Thai): https://ams.medsci.nu.ac.th/wp-content/uploads/2022/07/LAW2022/S13-%20กฎกระทรวง%20ความปลอดภัย%20อาชีวอนามัย%20และสภาพแวดล้อมในการทำงานฯ%202558.pdf
