Technical Guide

High-Voltage Substation — Monitoring of Modbus Network Analyzers

Monitor your high-voltage (HV) substations using Modbus TCP/RTU, Eziwan gateways, InfluxDB, and Grafana to track voltage, current, power, and harmonics.

Monitoring a high-voltage substation involves more than just checking whether a switchgear bay is energized. To understand power disturbances, anticipate faults, and correlate grid events with production outages, it is necessary to collect measurements from network analyzers, protective relays, meters, and Modbus devices. Eziwan centralizes this field data, transforms it into actionable metrics, and publishes it to InfluxDB for clear, historical visualization in Grafana.

The Problem

An industrial network administrator needs to collect measurements from 18 Schneider PM800 network analyzers via Modbus TCP and 12 Sepam protection relays via Modbus RTU at several high-voltage substations. The required data includes voltage, current, active and reactive power, power factor, frequency, harmonics, alarms, and certain protection statuses.

On paper, Modbus seems simple. In practice, monitoring a high-voltage substation quickly becomes complex.

  • The Modbus register tables for older devices are sometimes incomplete, scattered, or dependent on the firmware version.

  • Units, scale factors, and number formats vary by equipment family: signed integer, unsigned integer, 32-bit floating-point, high-word-first or low-word-first order.

  • Modbus TCP analyzers and Modbus RTU relays are not collected in the same way: topology, latency, timeout, and polling frequency must be adjusted accordingly.

  • Proprietary gateways often address a specific need, but they do not handle the addition of a new protocol, a new meter, or a business process change very well.

  • The lack of built-in transformations requires raw values to be converted on the cloud side, which creates an application layer that is costly to maintain.

  • The energy, maintenance, and production teams do not always share a common understanding of how to link an electrical disruption to a machine shutdown, a PLC alarm, or a decline in quality.

  • High-voltage substations are sometimes isolated, with limited or intermittent connectivity, which requires measurements to be buffered locally.

The challenge is therefore twofold: to ensure the reliability of Modbus data collection and to make the data immediately usable in an InfluxDB, Grafana, monitoring, or CMMS workflow.

Supervision Objective

In a high-voltage substation, electrical data is used to understand the actual behavior of the grid. It enables operators to track the load, detect imbalances, identify voltage sags, monitor harmonics, and analyze protection events.

Typical metrics to collect include:

  • Simple and compound voltages.

  • Currents per phase.

  • Active, reactive, and apparent power.

  • Power factor.

  • Frequency.

  • Active and reactive power.

  • Total Harmonic Distortion (THD) for voltage and current.

  • Harmonics by rank, depending on the analyzer’s capabilities.

  • Status of protection relays.

  • Alarms, faults, and trips.

  • Event timestamps, when available.

This data is most valuable when it is tracked over time and correlated with production events: line stoppages, drive failures, PLC restarts, protection trips, or drops in efficiency.

The target architecture places an Eziwan gateway in the substation or near the electrical area. It collects data from Modbus TCP devices over Ethernet, queries Modbus RTU devices over RS-485, applies the necessary transformations, and then publishes the time series to InfluxDB. Grafana provides dashboards for energy and power quality.

This architecture eliminates the need for multiple data collection software programs and makes electrical data available in a standard format that can be used by the energy, maintenance, and production teams.

Our Approach

Eziwan centralizes Modbus data collection, register processing, and transmission to monitoring tools. The gateway acts as an industrial data collector capable of communicating with existing equipment, diagnosing data exchanges, and standardizing data before publication.

  • Real-time diagnostic interface: View the raw values read from each Modbus register via the web interface, without the need for external tools.

  • Library of manufacturer templates: predefined configurations for common Modbus meters, analyzers, drives, and sensors, including those from Schneider, ABB, Siemens, and Danfoss.

  • Support for legacy equipment: Support for Modbus RTU, Modbus TCP, and Modbus ASCII scenarios, as well as gateways associated with historical Schneider environments, depending on the site’s architecture.

  • Onboard processing: conversion of raw register values to physical quantities, application of scaling factors, renaming of data points, site tags, and timestamping.

  • Publishing to InfluxDB: Writing metrics as time series ready for Grafana.

  • Local buffer: temporary storage of data in the event of a network outage, followed by data recovery once connectivity is restored.

  • Collector monitoring: equipment status, Modbus errors, response times, time-out frames, availability by equipment.

This approach integrates seamlessly with the Eziwan gateway, industrial connectivity architectures, and centralized monitoring via the Eziwan cloud.

Modbus TCP and Modbus RTU in a High-Voltage Substation

Modbus TCP and Modbus RTU are based on similar logic, but their operational constraints differ. In a mixed high-voltage substation, data collection must be adapted to each medium.

CriterionModbus TCPModbus RTU
SupportEthernet IPRS-485 or RS-232
AddressingIP address and portSlave address on serial bus
PerformanceFaster, depending on the networkDepends on baud rate and number of devices
DiagnosticsPing, TCP port, network logsBus analysis, timeout, collision, cabling
Common IssuesWrong port or firewallIncorrect termination, parity, address, polarity
Typical UsesRecent analyzers, gatewaysRelays, meters, and legacy equipment

On an RS-485 bus, the polling interval must be calculated carefully. Polling too many devices too frequently can cause timeouts, collisions, or unnecessary latency. On Modbus TCP, the network is more flexible, but it is still important to limit unnecessary requests and monitor application errors.

Mapping of Facilities

Before configuring data collection, you must create a clear inventory of the equipment. This helps prevent duplicates, misinterpreted records, and correlation errors.

FamilyExampleProtocolUseful data
Power AnalyzerSchneider PM800Modbus TCPVoltage, current, power, harmonics
Protection RelaySepamModbus RTUStatus, alarms, faults, electrical measurements
Energy meterModbus meterRTU or TCPActive and reactive energy, meter reading
Local PLCSubstation PLCTCPCell statuses, fault summary
Serial gatewayRS-485/IP converterTCP to RTUAccess to legacy equipment

This map must also indicate the site, the cell, the transformer, the power line, the IP address, the Modbus address, the baud rate, and the criticality.

Modbus Registers: From Raw to Physical

A Modbus register alone is not enough. To obtain a usable measurement, you need to know its type, unit, scale factor, and sometimes the word order.

Example of a transformation:

equipement:
nom: pm800_tgbt_01
protocole: modbus_tcp
ip: 192.168.20.41
port: 502

mesures:
- nom: tension_l1_l2
registre: 3020
fonction: holding_register
type: float32
word_order: high_low
unite: V
facteur: 1
tags:
site: usine_nord
poste: hta_1
depart: ligne_a

- nom: puissance_active_totale
registre: 3054
fonction: holding_register
type: int32
facteur: 0.001
unite: kW
tags:
site: usine_nord
poste: hta_1
depart: ligne_a

The benefit of on-board processing is that it avoids sending incomprehensible raw values to InfluxDB. The platform receives consistent physical quantities directly.

Real-Time Registry Diagnostics

The most time-consuming phase of a Modbus project is often the investigation. A documented register may be offset by one address depending on the convention used, a value may be encoded over two words, or a scaling factor may be applied within the equipment.

A diagnostic interface must allow you to check:

  • Equipment availability.

  • Response time.

  • The function code used.

  • The requested registry.

  • The raw value that was read.

  • The converted value.

  • Returned Modbus errors.

  • Timeouts and connection failures.

  • The actual polling frequency.

  • The timestamp of the last valid measurement.

This diagnostic process reduces the need for external tools and speeds up commissioning, especially when the equipment is old or poorly documented.

Data Pipeline to InfluxDB

InfluxDB is well-suited for industrial time series: repeated measurements, context tags, retention, aggregation, and visualization via Grafana. The quality of the data model is critical.

A properly modeled measurement must include a value, a timestamp, and relevant tags. For example: site, substation, transformer, feeder, equipment, phase, and unit.

Example of an InfluxDB data point

An InfluxDB write can represent an electrical measurement along with its industrial tags.

network_quality,site=north_plant,position=hta_1,equipment=pm800_tgbt_01,phase=l1_l2,voltage=401.7 1719403200000000000
network_quality,site=north_plant,position=hta_1,equipment=pm800_tgbt_01 active_power_kw=842.5 1719403200000000000
network_quality,site=north_plant,position=hta_1,equipment=pm800_tgbt_01 thd_voltage_pct=2.8 1719403200000000000

This structure then allows you to filter by site, position, equipment, or phase in Grafana, without having to create multiple databases or dashboards.

Grafana Dashboard for Network Quality

A Grafana dashboard should answer questions from field teams, not just display graphs. For a high-voltage substation, the useful views are:

  • Overview of the station: voltage, current, power, frequency, THD, alarms.

  • View by circuit: load, imbalance, active power, and reactive power.

  • Network quality monitoring: dips, surges, harmonics, frequency.

  • Protection view: SEPAM statuses, alarms, trips, faults.

  • Collection availability view: reachable devices, timeouts, latest measurements.

  • Production correlation view: electrical events and line outages.

  • Historical view: comparison by day, week, month, and reference period.

An effective dashboard should avoid cluttering the screen with unnecessary charts. The panels should guide decision-making: monitoring, diagnosing, correlating, or triggering an action.

Correlating Network Disruptions and Production Shutdowns

The value of the project becomes apparent when electrical data explains production events. A line that shuts down at 2:03 p.m. may be correlated with a voltage dip, a phase imbalance, a SEPAM alarm, or a spike in harmonics.

This correlation does not replace electrical analysis, but it provides a factual basis. It helps distinguish a machine failure from a power supply disturbance or an upstream grid event.

Collection Frequency and Prioritization

Not all measurements need to be collected at the same frequency. Too high a frequency overloads the Modbus bus, the gateway, and the database. Too low a frequency may cause an important event to be missed.

DataEstimated FrequencyComments
RMS voltage and current1 to 10 secondsDepending on criticality and equipment capacity
Active and reactive power5 to 30 secondsUseful for load monitoring
Cumulative energy1 to 5 minutesSlow variation
THD and harmonics10 to 60 secondsTo be adjusted as needed based on grid quality
Relay statuses1 to 5 secondsImportant for alarms
Protection eventsAs soon as availableHigh priority
Diagnostic data collection30 to 60 secondsEquipment availability

These values must be adjusted based on the bus, the equipment, the InfluxDB volume, and the expected level of detail.

Handling Timeouts and Missing Data

In electrical monitoring, the absence of data is itself a piece of information. It is important to distinguish between a true zero reading and a reading that is unavailable.

Best Practices:

  • Explicitly mark devices that have timed out.

  • Retain the timestamp of the last valid value.

  • Do not automatically replace missing data with zero.

  • Create alerts for missing data.

  • Monitor the error rate for each piece of equipment.

  • Use local caching when the network connection is unstable.

  • Play back the buffered data with their original timestamps.

This rigor prevents false conclusions in Grafana, particularly in power or energy charts.

Safety and Segmentation of the High-Voltage Substation

Electrical data is sensitive. It reveals the load, incidents, operating schedules, and sometimes the criticality of a line. Data collection must therefore be secure.

Recommended actions:

  • Isolate the high-voltage substation’s power system in a dedicated area.

  • Limit data traffic between the gateway and the Modbus devices.

  • Avoid any direct exposure of analyzers or relays to the Internet.

  • Encrypt outgoing data flows to the data platform.

  • Log connections and configuration changes.

  • Restrict Grafana access by role.

  • Document the equipment, ports, protocols, and permissions.

  • Establish a procedure for revoking service provider access.

This approach aligns with the OT/IT segmentation and least privilege principles detailed in the guides on industrial connectivity and secure access via the Eziwan gateway.

Manufacturer Templates

Templates significantly speed up commissioning. They describe the registers, types, scaling factors, units, and measurement names for known equipment.

A useful template should include:

  • The exact model or equipment family.

  • The version of the documentation used.

  • The registers that were read.

  • Data types.

  • Economies of scale.

  • The units.

  • Recommended tags.

  • Collection frequencies by measurement group.

  • Known limitations.

  • A field validation procedure.

Templates do not replace actual testing. Two devices from the same product family may have different options or versions.

Example of a multi-device configuration

poste:
nom: hta_usine_nord
timezone: Europe/Paris

collecte:
modbus_tcp:
- nom: pm800_depart_ligne_a
ip: 192.168.30.11
port: 502
template: schneider_pm800_qualite_reseau
intervalle: 5s
tags:
poste: hta_1
depart: ligne_a

- nom: pm800_depart_ligne_b
ip: 192.168.30.12
port: 502
template: schneider_pm800_qualite_reseau
intervalle: 5s
tags:
poste: hta_1
depart: ligne_b

modbus_rtu:
port: /dev/rs485-1
baudrate: 19200
parite: paire
stop_bits: 1
equipements:
- nom: sepam_cellule_arrivee
adresse: 1
template: schneider_sepam_etats_protection
intervalle: 2s
- nom: sepam_depart_moteurs
adresse: 2
template: schneider_sepam_etats_protection
intervalle: 2s

sortie:
influxdb:
bucket: qualite_reseau
measurement: poste_hta
buffer_local: true

This type of configuration makes deployment reproducible. It also facilitates auditing, since each collected data point is linked to a piece of equipment and a template.

Useful Alerts

Alerts must be actionable. An alert that is too vague creates noise; a contextualized alert helps with diagnosis.

Examples of relevant alerts:

  • Loss of communication with a PM800 analyzer.

  • Loss of communication with a Sepam relay.

  • Voltage dip below the defined threshold.

  • Current imbalance between phases.

  • THD voltage or current exceeds the monitoring threshold.

  • Protection trip or alarm.

  • Lack of data on a critical item.

  • Increase in the Modbus timeout rate.

  • High reactive power over an extended period.

  • Unusual load discrepancy between departures.

The thresholds must be approved by the electrical teams. We must not allow the monitoring system to become a source of unqualified alerts.

Pre-Production Checklist

CheckQuestionPriority
InventoryAre all analyzers and relays identified?High
DocumentationAre the Modbus tables validated?High
InterfacesAre TCP, RTU, baud rate, and addresses confirmed?High
TemplatesHave conversions and units been tested?High
TimestampingAre the time zone and time synchronization defined?High
InfluxDBAre the tags and measurements consistent?High
GrafanaDo the dashboards meet business needs?High
AlertsHave the thresholds been validated by the electrical team?High
SecurityAre data streams segmented and encrypted?High
BufferAre temporary network outages handled?Medium

This checklist helps you avoid common mistakes: incorrectly converted data types, inconsistent units, measurements without timestamps, or dashboards that are unusable.

Common Mistakes to Avoid

Collecting Too Many Records from the Start

It’s tempting to collect everything. But collecting too much data overloads the buses, complicates the dashboards, and increases storage requirements. It’s better to start with the metrics that are useful, then expand as needed.

Ignore scale factors

A raw value can represent volts, kilovolts, amperes, tenths, or thousandths of a unit. Without a scale factor, Grafana charts may be inaccurate even though they appear consistent.

Confusing null data with missing data

A power rating of zero and a measurement that is unavailable are not the same thing. Timeouts must be handled explicitly.

Neglecting the quality of the RS-485 bus

A bus that is improperly terminated, too long, improperly polarized, or disrupted can cause intermittent errors that are difficult to diagnose. Communication diagnostics are just as important as software configuration.

Building Grafana Without Business Logic

A good dashboard should help you make decisions. The charts should be organized by position, departure, equipment, and event, with clear thresholds and useful filters.

Performance Metrics

To make the most of monitoring over the long term, it is necessary to track both the electrical data and the quality of data collection.

IndicatorPurpose
Average, min, and max voltageDetect sags, surges, and instabilities
Current per phaseIdentify imbalance and overload
Active powerMonitor load and consumption
Reactive powerMonitor compensation and power factor
Voltage and current THDAnalyze grid quality
Sepam alarmsCorrelate faults and events
Equipment availabilityVerify data collection
Modbus timeout rateDetect bus or network issues
Last valid measurementIdentify silent equipment
Number of correlated eventsLink the power grid and generation

These indicators enable a phased approach: monitoring, diagnostics, quality analysis, and then energy optimization.

How Eziwan Simplifies HTA Oversight

NeedEziwan’s ResponseBenefit
Collect Modbus TCPNetwork analyzer pollingCentralized electrical measurements
Collect Modbus RTURS-485 interfaceIntegration of relays and legacy equipment
Understand registersReal-time diagnosticsFaster commissioning
Standardize dataEmbedded transformationsActionable physical values
Accelerate projectsManufacturer templatesLess manual configuration
Log dataInfluxDB publishingGrafana-ready time series
MonitorLogs and alertsSimpler diagnostics
SecureTunnels and segmentationNo direct exposure of equipment

Eziwan transforms a scattered Modbus data collection into a usable industrial data pipeline that meets network quality, maintenance, and production requirements.

Conclusion

Monitoring Modbus network analyzers in a high-voltage substation enables a shift from a one-time assessment to continuous analysis of power quality. By collecting data from PM800s, Sepams, and other Modbus devices, and then storing the measurements in InfluxDB and Grafana, teams can correlate grid disturbances with production outages, diagnose issues more quickly, and better prioritize actions.

Eziwan provides the necessary building blocks: Modbus TCP and RTU data collection, register diagnostics, manufacturer templates, embedded transformations, local buffering, InfluxDB publishing, alerts, and data stream security. For industrial sites, this approach makes HTA monitoring more reliable, easier to understand, and directly actionable by field teams.

Further Reading

Frequently Asked Questions

You might also like