Factory solar power monitoring should do more than display an inverter production curve. It must connect PV generation, factory load, grid import and export, alarms, relevant irradiance and weather, contracts and time synchronisation into operational evidence that management, engineering, IT and procurement can reproduce. This guide shows how a factory in Thailand or ASEAN can specify that evidence through an RFP, PoC, FAT/SAT and operational acceptance.
Turn factory solar power monitoring into an evidence system
An inverter or PCS portal is useful for checking the generating asset. It rarely answers all the questions the factory must manage: how much grid electricity was displaced at the same time, whether an import peak occurred when clouds passed, why an alarm lasted, and whether monthly figures reconcile to the utility billing period.
The systems observe different boundaries. An inverter sees the PV side; the utility or incoming meter sees the point of common coupling; submeters see only selected factory loads. Each can be correct in isolation while the combined report is wrong because clocks, signs, intervals, CT/PT ratios or missing-data rules differ.
Use an explicit energy balance as the governing model:
PV generation + grid import = factory load + grid export + losses or unmetered energy inside the boundary
This is an accounting relationship, not a promise that every sample will equal zero residual. Transformer and cable losses, PV auxiliaries, different metering locations, batteries, clock drift and missing values create differences. The acceptance objective is to explain the residual within agreed limits, not to hide it.
IEA PVPS Trends 2025 reports that global cumulative PV capacity exceeded 2,260 GW at the end of 2024. That global figure does not predict the yield or payback of an individual factory. It shows why long-term performance evidence is becoming as important as installation itself.
Five gaps left by an inverter portal alone
- Generation is visible, but factory load at the same time is missing, so self-consumption cannot be demonstrated.
- Import/export signs and meter locations are unclear, causing reverse-flow errors.
- PCS alarms arrive, but start, acknowledgement, restoration and impact are not one auditable record.
- Monthly output is visible, but irradiance, soiling, temperature, curtailment and failure are not separated.
- Portal periods and multipliers do not match the utility bill, so finance cannot reproduce the result.

Seven data groups for factory solar generation visualisation
A solar generation visualisation RFP should specify evidence and decisions rather than screen names.
| Data group | Typical source | Question it must answer |
|---|---|---|
| PV generation | PCS, AC combiner or revenue-grade meter as required | When and how much did the PV system generate? |
| Factory load | Incoming, transformer, line and utility submeters | What consumed electricity at the same time? |
| Grid import/export | Bidirectional meter at the defined coupling point | Can purchased and exported energy be explained with one sign convention? |
| Alarm and state | PCS, protection devices, gateway | When did an event start, who responded and when was service restored? |
| Weather/irradiance | Selected irradiance and temperature sensors or a documented service | Was reduced production caused by resource, condition or equipment? |
| Contract/configuration | Tariff, demand contract, interconnection, SLA, CT/PT ratios | Which rule and configuration governed the calculation? |
| Quality/audit | Clock, gaps, calibration and changes | Can another person reproduce the conclusion later? |
IEC 61724-1:2021 covers terminology, equipment and methods for PV performance monitoring and defines monitoring-system classes. Treat it as a framework for choosing monitoring quality according to purpose and risk, not as a sentence that automatically defines every factory sensor. The RFP should still state quantities, accuracy, sampling, retention, gap handling and maintenance.
Make the metering-boundary drawing the first deliverable
Add all meters to the single-line diagram and identify whether each observes the PV side, the load side or the grid side. When a high-voltage incoming meter and a low-voltage PV meter sit on opposite sides of a transformer, losses and location differences matter. A low-voltage system can also contain PV auxiliaries or unmetered branches.
The drawing and meter register should include the coupling point, transformer, PV switchboard, PCS units, principal loads and any battery; tag, voltage level, direction, positive sign, CT/PT ratio, unit and value type; the path through gateway and cloud; the time source; and the responsibility boundary between utility and factory records. Operating modes such as non-export, self-consumption and excess export must be described as intended or approved states, never assumed permissions.
Connect PCS monitoring to factory maintenance
PCS monitoring in a factory is not complete when communication succeeds. It must connect instantaneous power, cumulative energy, DC/AC state, events, curtailment and communication health to the plant’s response process. Preserve the original manufacturer event code and map it to a common factory classification.
| Common class | Example | Required response |
|---|---|---|
| Generation stopped | PCS trip, protection operation, grid-condition stop | Severity, contact, site safety check and restoration approval |
| Performance degraded | String deviation, thermal condition, suspected limitation | Compare irradiance and history, then dispatch inspection |
| Communication failed | Gateway outage, frozen tag, delayed portal | Distinguish data outage from generation outage |
| Data quality failed | Gap, fixed value, clock drift, counter reset | Flag the period; do not silently replace it with zero |
| Maintenance due | Inspection, calibration, firmware or certificate | Use controlled outage and change-management procedures |
An SLA should separate clocks for detection, notification, acknowledgement, first diagnosis, site attendance, temporary restoration and permanent correction. “24/7 monitoring” alone cannot be tested. Generation stops and communications failures also require different priority. Night-time zero output should not trigger the same logic as a daytime trip.
Use only the weather and irradiance evidence the decision needs
The IEA PVPS analytical monitoring guidance discusses instruments, measured quantities, losses and performance analysis. A factory should select sensors based on the faults, guarantees and O&M decisions it must support. An irradiance sensor with wrong orientation, shading, soiling, clock or calibration can create a misleading baseline.
Local sensors and external weather services can both be valid in the right context. State the source, spatial and temporal resolution, delay, maintenance and gap policy. Do not mix them without traceability. More sensors are not automatically better; maintainable evidence is better.
Solar self-consumption monitoring needs one time window
The common error in solar self-consumption monitoring is subtracting monthly utility energy from a daily or portal total that uses a different boundary. When no battery is present and boundaries align, the concepts are:
- Self-consumed PV = PV generation − grid export.
- Self-consumption ratio = self-consumed PV ÷ PV generation.
- Solar fraction = self-consumed PV ÷ factory load.
Before implementing those equations, define export sign, CT/PT ratio, clock, transformer losses, counter resets and missing periods. If a battery exists, charging, discharge, losses and state transitions must be included. Specify handling of night-time, shutdowns, outages and zero denominators.
Combine quantity, performance, quality and action
Generation alone rewards a sunny day even if the factory is closed. Self-consumption ratio alone can make an undersized PV system look excellent. Use a balanced set.
| View | Example KPI | Interpretation guardrail |
|---|---|---|
| Energy | PV, import, export, self-consumed energy | Same period, boundary and unit |
| Performance | Availability and expected-versus-actual indicator | Separate irradiance, temperature, curtailment and outage |
| Demand | Peak within the tariff demand interval | Confirm tariff and contractual interval |
| Data quality | Gap rate, clock drift, unexplained residual | Never improve appearance by silent filling |
| Response | Acknowledgement/restoration time and recurrence | Define owner and exclusions |
For a broader plant specification, see our Thailand factory energy monitoring RFP and FAT/SAT guide. It shows how to connect PV with billing, major loads, compressed air, water and BESS under one acceptance model.

Demand control must not rely on a PV forecast alone
Demand control monitors the demand interval defined by the tariff or contract and manages controllable loads before a peak. PV may reduce daytime import, but passing clouds, a PCS trip and simultaneous production start-up can create a fast import rise. The larger the PV contribution, the more important it is to compare PV ramp rate with the response time of permitted loads.
Separate five layers in the specification:
- Monitoring: current demand, interval-end forecast, PV and controllable load.
- Notification: thresholds, recipient, acknowledgement and escalation.
- Recommendation: candidate actions constrained by production and safety.
- Automatic control: only approved loads, with interlocks, minimum run times, release conditions and manual override.
- Verification: before/after values, commands, production impact and reasons for inaction.
A dashboard does not authorise safe load shedding. Chillers, compressors, furnaces, pumps and chargers carry process, safety, quality and equipment-life constraints. Electrical, production, facilities and safety owners must approve automation and define the fail-safe state for loss of communication.
Thailand: voltage level, interconnection and the 250 kWp question
Requirements cannot be inferred from PV capacity alone. The relevant distributor, high- or low-voltage connection, coupling point, export status, self-consumption arrangement, reverse-power protection, contract and project timing all matter.
PEA’s official grid-connection information page provides entry points to connection rules, point checks and applications for self-use generation without electricity sales. The PEA PPIM portal displays routes for the 2026 rooftop excess-power programme, connection checks, VSPP and self-use/non-sale generation. These pages show that pathways exist; they do not prove that every factory qualifies, can export or can connect.
PEA’s August 2026 announcement on installer and equipment registration discusses installers, modules, inverters, batteries, installation quality and inverter testing before grid connection. Its language strongly addresses public/people-sector safety and access. A factory may use registration as one due-diligence input, but it must not treat it as automatic approval or compliance for an industrial project.
How to state the Power Quality Meter issue above 250 kWp
The ERC Rooftop PV System page contains a FAQ for the 2013 programme stating that systems with installed module capacity above 250 kWp must provide a Power Quality Meter at the connection point meeting distributor requirements. Because this is a 2013 programme page, do not claim automatic application to every 2026 industrial self-consumption or export project.
The RFP should require the bidder to confirm current applicability with ERC, the responsible PEA or MEA office and qualified electrical/legal professionals, and to submit the governing document, version and approval record. Do not infer that systems below 250 kWp need no project-specific review. Protection, metering, power quality, reverse flow and applications remain project dependent at both high and low voltage.
PEA’s 1 September 2026 Solar Rooftop energy management project with the Provincial Waterworks Authority is a current example of integrating solar assets with energy management at organisational facilities. It is not evidence for a factory’s regulatory status, cost, yield or payback.
Write testable solar monitoring RFP deliverables
Do not begin with “cloud monitoring package” or a count of dashboards. Begin with deliverables that can be accepted.
Data dictionary and meter register
For every tag, define name, description, asset, metering point, direction, unit, multiplier, type, expected range, sample, aggregation, retention, time, missing-value rule, quality flag and source. Keep the source tag and transformation even when manufacturer names are normalised.
Calculation and reconciliation specification
State formulas for self-consumption, import/export, load, generation, demand and performance; include timezone, billing cut-off, rounding, counter resets and missing data. The dashboard, API and CSV must produce the same result for the same interval.
Responsibility boundary and SLA
Use a RACI covering EPC, PCS manufacturer, monitoring vendor, factory facilities, IT, distributor and O&M. Define who diagnoses the asset, network, cloud, account and sensor. Include administrator rights, configuration backups, API access, raw-data return and migration at contract end.
Cybersecurity and continuity
Avoid creating an unnecessary path into the PCS control network. Specify read-only access where possible, segmentation, outbound paths, authentication, certificates, remote access, logging, patching and backup. Protection and safe generation operation should not depend on public cloud availability.
Time synchronisation and change records
Define a common source and timezone policy for PCS, meters, gateways, servers and cloud. Test drift, correction, delayed data, counter reset and restart. Record who changed a setting, when, old and new values, reason, approval and affected reports.
When PV data is used for decarbonisation reporting, see our factory CO2 emissions visualisation guide for Thailand. Emission-factor versions, market- versus location-based reporting, instruments and double-counting controls require a separate evidence chain.
Use a PoC to test hypotheses, not a temporary showroom
A PoC should test the highest-risk decisions within a small, real boundary. Suitable acceptance hypotheses include distinguishing cloud-driven reduction from a PCS event, reconciling 15-minute PV/import/load values, and identifying a communications outage without labelling it as generation failure.
| Phase | Work | Exit evidence |
|---|---|---|
| Days 0–30 | Establish boundary, tags, clocks, links and baseline | Values, signs, multipliers and time reconcile to field instruments |
| Days 31–60 | Test alarms, gaps, irradiance/load variation and reports | Events reach the correct owner with explainable evidence |
| Days 61–90 | Exercise monthly reconciliation, SLA, recovery and handover | Factory operators can run the process without hidden vendor dependency |
Ninety days is an example, not a universal requirement. Choose enough time to observe relevant operating and seasonal conditions. Agree pass/fail evidence and the treatment of unmet criteria before starting.

FAT, SAT and operational acceptance
FAT verifies calculations, displays, alarms, roles, exports and failure behaviour in the supplier environment. SAT verifies CT/PT wiring, polarity, meter direction, network, clocks and connection points in the real factory. Passing FAT does not prove field metering.
FAT tests
- Feed known values and reconcile PV, import/export, self-consumption and demand.
- Inject gaps, frozen values, time reversal, duplicates and counter resets; inspect quality flags.
- Distinguish PCS stop, communications loss, sunset and curtailment.
- Verify role-based changes, audit logs and API/CSV completeness.
- Confirm separate SLA timestamps for detection, notification, acknowledgement and restoration.
SAT tests
- Compare CT polarity, phase sequence, CT/PT ratios and direction with field instruments.
- Confirm actual import/export signs under approved operating conditions.
- Compare PCS, incoming and load clocks against the same event.
- Interrupt connectivity and restart the gateway; verify buffering without gaps or duplicates.
- Never bypass protection for a monitoring test; qualified personnel must execute approved procedures.
Operational handover
Transfer as-built drawings, single line, meter register, data dictionary, formulas, alarm matrix, SOPs, backups, administrator rights, calibration evidence, configuration, licences, contacts, SLA, training, FAT/SAT records and open issues. Demonstrate extraction of raw data and configuration before accepting a vendor cloud dependency.
ISO 50001 provides a framework for managing and continually improving energy performance. Certification is possible but not mandatory. Its value here is the management cycle: connect PV evidence to objectives, responsibilities, review, corrective action and improvement instead of leaving it as an engineering-only screen.
Acceptance questions by function
| Function | Main question | Evidence |
|---|---|---|
| Management/finance | Can purchased energy, risk and contractual effect be explained? | Monthly reconciliation, exceptions and approved KPIs |
| Plant/production | Will peak action preserve safety, output and quality? | Control constraints, actions and impact records |
| Facilities/energy | Can faults be detected and safely restored? | Alarm, SOP, calibration, maintenance and SAT evidence |
| IT/OT | Can connections, access, logs and recovery be governed? | Architecture, accounts, audit and recovery tests |
| Procurement/legal | Are outcome, SLA, data ownership and exit clear? | Compliance matrix, RACI, contract and export test |
Turn the monthly review from reporting into approved action
After go-live, exception handling matters more than the page count of a report. A monthly review should follow a fixed order: data quality, energy balance, performance and operating context, then corrective action. First inspect gaps, clock drift, counter resets, CT/PT changes and unapproved manual edits. Mark periods with material quality problems as provisional instead of blending them into a final report.
Next, recalculate PV, load and import/export for the same window and investigate days with an abnormal residual. Consider wiring or configuration changes, a new load, transformer switching and gateway updates as well as equipment faults. For performance, do not conclude from a simple year-on-year comparison. Record irradiance, temperature, operating days, planned outages, curtailment, cleaning and construction. If resource-data quality is inadequate, retain “cause not confirmed” rather than forcing a normal/abnormal label.
Every corrective action needs an owner, due date, required outage, expected result and completion evidence. “Check the inverter” is not testable; “compare the original PCS event log with the local display and determine whether the event was a communications or generation outage” is. When a setting or formula changes, approve the period to be recalculated and preserve before/after results.
Review the monitoring scope quarterly or half-yearly. Remove tags and notifications that support no decision, add evidence that was missing during diagnosis, and inspect SLA quality rather than a single compliance percentage. Different retention periods may be appropriate for high-frequency samples, demand intervals, monthly approved values, original alarms, tickets and configuration backups. If raw data is deleted after aggregation, retain when and how aggregation occurred. A cloud export is not a backup until it has been restored elsewhere with tags, units, quality flags and timezone intact.
FAQ about factory solar power monitoring
Is an inverter portal enough for solar generation visualisation?
It may be enough for basic PCS maintenance. It is not enough to demonstrate self-consumption, grid import/export, tariff demand and factory load unless these are integrated with the same boundary and clock.
What should solar self-consumption monitoring measure?
At minimum, PV generation, factory load and bidirectional grid flow. Add battery charge/discharge if applicable. Define sign, CT/PT ratio, time window, losses and gap rules so that residual energy can be explained.
How should PCS monitoring connect to maintenance?
Retain the original event and map it to factory classes for stop, degradation, communications, data quality and maintenance. Link detection, acknowledgement, field response, restoration, cause and recurrence prevention in one record.
Does solar eliminate demand control?
No. Clouds, PCS trips and coincident process starts can raise import quickly. Design monitoring, notification, recommendation and approved automatic control around the actual tariff interval and process constraints.
Is a Power Quality Meter mandatory above 250 kWp?
The ERC’s 2013 rooftop programme page contains that statement, but its application to a specific 2026 factory must be confirmed. Consult ERC, the responsible PEA/MEA office and qualified professionals using the project’s voltage, capacity, connection point and operating/export mode. Projects below 250 kWp also require project-specific review.
What does the system cost and what is the payback?
It depends on existing meters, point count, communications, weather evidence, retention, SLA, cybersecurity and field work. Do not compare unsupported package prices or payback claims. Define common evidence and acceptance conditions, then request comparable bids.
Conclusion: make the generation curve common operational evidence
Factory solar power monitoring is not another portal. It integrates PV generation, factory load, grid import/export, alarms, necessary irradiance and weather, contracts/SLA and time synchronisation under one boundary. Define data, calculations and responsibility in the RFP; test hypotheses in the PoC; test abnormal cases in FAT/SAT; then transfer access, procedures and evidence to factory operations. Confirm Thailand-specific high/low-voltage, interconnection, self-consumption, export and 250 kWp-related requirements with ERC, PEA/MEA and qualified professionals for the actual project.
You can consult TOMAS TECH while the metering boundary and RFP are still being shaped. To turn existing PCS, incoming-meter, load and contract information into a PoC and acceptance plan, contact our team.