Online Monitoring of Wind Turbines: Database Servers and Live Displays

Sunday, shortly before three in the morning. The turbine reports a fault, shuts down and does not reset itself. The wind is blowing hard, but the turbine is at a standstill. The message sits in the controller’s memory, but nobody reads it. On Monday morning the downtime is noticed, because a neighbour calls or because the direct marketer’s statement looks odd. By then, many hours of good wind have passed.

This is exactly what online monitoring is for. It reads the wind turbine’s controller continuously, stores the values in a database, shows the status on a live display and raises an alarm when something is wrong. That sounds obvious. In practice, however, we often see gaps at single turbines and small wind farms: data that is held in the controller for only a few weeks, access that only the manufacturer has, or alarms that go to a mailbox nobody reads any more.

We have operated our own wind turbines since 1998, and whatever we develop runs in our own fleet first. In this article we show what manufacturer-independent online monitoring should deliver, how a fully automatic database server records operating data without gaps, what live displays show and which key figures you can use to assess the operation of your turbine. At the end you will find typical mistakes and answers to frequently asked questions.

What manufacturer-independent online monitoring delivers

Online monitoring is more than remote access to the controller. It consists of three parts that have to work together: acquisition at the turbine, storage on a database server and analysis, meaning display, key figures and alerting. If one of these parts is missing, you have either data without any use or displays without a memory.

Manufacturer-independent means that the system works regardless of which manufacturer built the turbine. For operators of mixed wind farms or older turbines, this is often the decisive point, because the manufacturer’s monitoring portal usually covers only its own models.

Which values should be recorded

Which values a controller supplies depends on the turbine type and the controller generation. For reliable monitoring you need at least these groups:

• Power: active power and, where required, reactive power and power factor.
• Wind: wind speed and wind direction at the nacelle anemometer, plus the nacelle position.
• Rotational speeds: rotor and generator speed and, on turbines with blade pitch control, also the pitch angle, i.e. the angle of attack of the rotor blades.
• Operating states: operation, standby, fault, maintenance, curtailment, icing, shadow flicker or bat shutdown.
• Fault messages: code, plain text, time of occurrence and time of acknowledgement.
• Meter readings: energy generated, own consumption and operating hours.
• Temperatures: gearbox, generator, bearings, nacelle and ambient, where the controller outputs them.

One system for mixed wind farms

In many small wind farms, turbines from different manufacturers and years of construction stand side by side. Each has its own controller, its own status codes and its own access. Manufacturer-independent monitoring brings these sources together in a shared database. You see all turbines on one interface, with consistent units and comparable key figures.

This has a second advantage: your operating data no longer sits only with the manufacturer or the maintenance provider, but in a system that you access yourself. More on this in the section on data sovereignty.

The database server: recording operating data without gaps

The controller of a wind turbine is built to control, not to archive. Many older controllers keep averages and messages for only a limited period and then overwrite the oldest entries. If you do not collect the data regularly, you lose it.

A fully automatic database server takes over exactly this task. It polls the controller at fixed intervals, writes the values with a timestamp into a database and fills in gaps as soon as an interrupted connection is restored, provided the controller still holds the values. Nobody has to trigger anything.

10-minute averages, one-second values and event log

In the wind industry, the 10-minute average has become established as the standard for operating data. The FGW’s Technical Guideline TR 10 on site quality also works with what it calls 10-minute data. It permits any averaging interval of the SCADA system between five and fifteen minutes and explicitly distinguishes this continuous time series from the Event/Status Log, the event log with the start and end time of every message (FGW TR 10 Rev. 2, preview). SCADA stands for Supervisory Control and Data Acquisition, i.e. the system for monitoring, control and data acquisition.

For monitoring you generally need all three types of data, but not each in the same depth:

How much storage you should plan for

The storage requirement depends almost entirely on whether you record one-second values permanently. A rough estimate that we calculated for a single turbine with 50 measurement channels:

1. 10-minute values: 50 channels times 4 statistics (mean, minimum, maximum, standard deviation) times 144 intervals per day give 28,800 values a day, around 10.5 million a year. At 8 bytes per value, that is roughly 84 MB of raw data. With timestamps and indexes it stays well below one gigabyte per year.
2. One-second values: 50 channels times 86,400 seconds give 4.32 million values a day, around 1.58 billion a year. That is roughly 12.6 GB of raw data and, with management data, quickly a double-digit number of gigabytes per turbine and year.
3. Event log: usually negligible, as long as the turbine is not stuck in a fault loop generating thousands of messages a day.

For a small wind farm, even one-second values are no problem with today’s hardware. The real question is a different one: how long should the data remain available, and how is it backed up? A proven approach is to keep 10-minute values and events permanently and to record one-second values only for certain channels or after an event.

Database and backup

A database is only as good as its backup. Hard drives are wear parts, and a server in a substation or at the tower base works under different conditions from a computer in an office: temperature fluctuations, dust, voltage dips. Plan the backup from the very beginning:

• Mirrored storage media in the server, so that a single hard drive failure does not cost any data.
• Regular copy to a second location, for example to a server in the office or to the cloud.
• Tested restore: a backup that has never been restored is an assumption, not proof.
• Monitoring of the server itself: storage space, temperature and the last successful poll also belong in the alerting.

Live displays in the office and on site

The database is the memory, the live display the face of the monitoring. It shows the current status at a glance, without anyone having to start an analysis. Typical locations are the office of the technical operations manager, the meeting room of an energy cooperative or a screen on site, for example in a grid transfer station.

What a good display shows

A display has to answer one question within three seconds: is everything running? Everything else is extra. This layout has proven itself:

1. Status per turbine with an unambiguous colour or symbol: operation, standby, fault, maintenance, curtailment.
2. Current power per turbine and as a total, plus wind speed and wind direction.
3. Daily and monthly yield, so that laypeople can also see what the turbine has achieved today.
4. Open fault messages in plain text with the time.
5. Time of the last update, so that a frozen display is not mistaken for normal operation.

The last point is often forgotten. A display that has been showing the same values for two days because the connection has dropped lulls everyone involved into a false sense of security.

Example: the terminal replicas on our website

How different the user interfaces of wind turbines look is something we show on our home page in the section on database servers and live displays. There you will find terminal replicas in the style of Vestas RCS, Mita, Tac1 and Enercon displays, as well as a DV overview that combines several turbines. The values shown come from a wind simulation and are not data from real turbines.

The replicas illustrate one point: every controller generation presents states, messages and measured values differently. Manufacturer-independent monitoring has to read all these formats and translate them into a common language before a display or a key figure can be produced from them.

Key figures: how to assess the operation of your turbine

Raw data alone says little. Only key figures make it visible whether a turbine is doing what it should. For single turbines and small wind farms, five key figures are enough if you calculate them properly and read them correctly.

Availability: time-based and production-based

Availability is the most frequently cited and at the same time the most misunderstood key figure. The international standard IEC 61400-26-1:2019 describes an information model from which both time-based and production-based availability can be derived, for a single turbine as well as for an entire wind farm. At the heart of the model are categories into which every period of time and every quantity of energy is classified, with fixed rules on which category takes precedence when they overlap.

For you as an operator, the difference is what matters most. Take a turbine that is at a standstill for only a few hours a month, but precisely during the windiest hours. The time-based availability looks excellent. The production-based availability, by contrast, shows how much yield was actually lost. Always look at both values, therefore, and check carefully in maintenance and service contracts which definition a guaranteed availability is based on.

Capacity factor and power curve

The capacity factor puts the energy generated in relation to what the turbine would have generated at uninterrupted rated power. Closely related are the full-load hours: annual energy divided by rated power. Both values fluctuate strongly with the wind year and are therefore mainly suitable for comparisons over several years or with neighbouring turbines.

The power curve is more meaningful for the technical condition. To obtain it, we sort the 10-minute values into bins by wind speed and calculate the average power for each bin. If the measured curve stays below the manufacturer’s power curve over a longer period, this may indicate misadjusted blade angles, a misaligned wind sensor, soiled rotor blades or an active power limitation.

An important point: the anemometer on the nacelle measures behind the rotor, in air that is already turbulent. A power curve derived from operating data is therefore no substitute for a power curve measurement in accordance with a guideline, such as the one described in FGW guideline TR 2. For comparing the same turbine over time, however, it is a very good tool, because the measurement error has a constant effect.

Downtime analysis and yield comparison

The downtime analysis answers the question of why a turbine was not running. To do this, we add up the downtimes from the event log by cause: fault, maintenance, grid, curtailment by the direct marketer, shutdowns for bird, bat or ice protection. Sorted by duration, it quickly becomes clear whether a few large or many small outages are costing yield. Each calls for different measures.

In a yield comparison, you place your turbine alongside neighbouring turbines of a similar type and alongside wind data from the same region. If only your turbine falls short, the problem lies with the turbine. If they all fall short, it was a weak wind year. Without this comparison, operators and service providers often spend months debating assumptions.

Fault detection and alerting

Back to Sunday at three in the morning. No monitoring system can prevent the fault itself. What matters is how quickly someone finds out about it and whether that someone is able to act.

From status code to understandable message

Controllers report faults as codes, often with a short, manufacturer-specific text. Good monitoring translates these codes into understandable messages and assigns them to categories: does someone have to intervene, does the turbine reset itself, or is it merely a notice? With the state-event-cause key ZEUS (Zustand-Ereignis-Ursachen-Schlüssel) in its guideline TR 7, the FGW has published a cross-manufacturer system for the states and events of generating units. Even if you do not adopt it in full, consistent classification helps with later analysis.

In addition to genuine fault messages, indirect anomalies should also be able to trigger an alarm:

• No power, although the wind is above the cut-in speed and no curtailment is active.
• Rising gearbox or bearing temperatures at the same power output.
• No new data for a defined period, i.e. a loss of connection.
• Clear deviation from the neighbouring turbine in the same wind.

Escalation and on-call duty

Common channels for alarms are e-mail, text message and push notifications from an app. More important than the channel is the chain behind it. A tiered escalation has proven itself:

1. Level 1: the message goes to the person on call, for example the technical operations manager.
2. Level 2: if it is not acknowledged within a defined time, it is also sent to a deputy.
3. Level 3: if it still remains open, the service company or the operator is informed directly.
4. Documentation: who acknowledged it when and what action was taken is stored with the message.

Clarify in advance who can be reached at night and at weekends and what authority that person has. May they acknowledge and restart the turbine remotely? Must they always call the service company for certain codes? These rules belong in writing in the technical operations management, not in the memory of individual people.

Remote control belongs with the live display: turbines can be started, stopped and controlled remotely, but only while watching the current values – you should only control what you can see. Cameras are a separate matter: they show the turbine and the site live. Both shorten the response time, but do not replace checking whether a restart is permitted at all after a particular fault.

Remote administration during business hours

Two things are often confused: the automatic monitoring of the turbine and the support of the monitoring system itself. The monitoring runs around the clock: it detects faults, evaluates operating data and raises alerts. Remote administration of the system, meaning maintenance, adjustments and support with questions about the software, is handled by our office during normal business hours.

For you, this means that on-call duty for interventions at the turbine outside business hours is something you organise together with your technical operations management and your service company. The monitoring ensures that the right person receives the message in good time.

Data sovereignty and export

Operating data is an asset. You need it for expert reports, for continued operation after the end of feed-in subsidies, for discussions with insurers and service providers and when selling a turbine. Nevertheless, many operators do not know exactly where their data is stored and whether they can get all of it out.

CSV export and interfaces

The simplest and most robust form is the CSV export, a text file with columns that any spreadsheet program can open. Make sure that timestamps are unambiguous, i.e. with a time zone or in UTC, and that units are given in the header row. Otherwise every analysis begins with guesswork around the clock change.

Ongoing exchange with operations management software or the direct marketer requires interfaces. For communication for the monitoring and control of wind farms, the IEC 61400-25 series of standards provides an international framework. It describes information models for wind farms, a model for information exchange and the mapping to standard communication profiles based on the client-server principle. Many older controllers do not speak this standard, so the monitoring translates between the manufacturer-specific protocol and the outside world.

With direct marketing (Direktvermarktung) via EMSYS, the connection includes the data interface to the direct marketer and the implementation of schedules and curtailments at the turbine. If the data is stored in your own database, you can later reconcile curtailments with the statements.

What you should clarify in advance

1. Who owns the raw data, and where is it stored?
2. Can you export the data in full at any time, not just as a monthly report?
3. How long are 10-minute values, events and, where applicable, one-second values retained?
4. What happens to the data when the contract with a service provider ends?
5. Who else may access it, for example the service company, expert assessors or the direct marketer, and how is this logged?

Retrofitting older turbines

Turbines from the 1990s and early 2000s in particular are often solid machines with outdated data connections: analogue modems, serial interfaces, software for operating systems that hardly anyone still runs. Typical turbines of this generation come from manufacturers such as Vestas, Enercon, Südwind, DeWind or Tacke. A retrofit is particularly worthwhile here, because the data has usually not been recorded at all so far, or only with gaps.

Connection via the controller interface

The preferred route is via the controller’s existing interface. Many controllers have a service or remote monitoring port through which measured values, states and messages can be read out. A computer at the tower base or in the grid transfer station handles the protocol, buffers the data and sends it to the database server via mobile network (4G/5G) or fibre optic cable.

The advantage: you receive exactly the values and status codes that the controller itself uses. The challenge lies in the documentation. Protocol descriptions are not always available for older controllers, and some codes first have to be mapped through observation.

Connection via dedicated measurement equipment

Where the controller does not offer a usable interface, additional measurement equipment helps: a power meter at the generator output, a meter reading head, a separate wind sensor or tapping status signals such as operation and fault via relay contacts. This provides less detail than the controller itself, but is often sufficient for yield monitoring, availability and fault alarms.

In practice we combine both routes, depending on what the turbine offers. Whether and how your turbine can be connected is something we determine during the technical clarification. In the process, we bring existing remote monitoring systems up to date: connection to the internet, transfer to the cloud and access from anywhere.

Process and scope of delivery

Thomas Wilmes supports every connection personally and brings in professional service companies for installation and service. A project for a database server, live display or online monitoring follows fixed steps with us.

Five steps from order to handover

1. Order confirmation and technical clarification: after this, we determine the expected delivery date. Quotations remain valid for three months.
2. Design and manufacturing of the system within the agreed scope of delivery.
3. Readiness for shipment and transport to the site.
4. Installation and commissioning on site, mechanical and electrical.
5. Handover, training and documentation: training on the system takes place during the commissioning phase. Final acceptance follows no later than two weeks after the start of commercial use.

The client is responsible for providing all necessary construction work, such as supply lines and telephone connections, as well as the power supply. Clarify these points early on, because a missing supply line delays commissioning faster than any technical question. After handover, we offer maintenance and spare parts as required.

Warranty and wear parts

We provide a 24-month warranty on parts, workmanship, design, electrics and controls. It begins on the day of commercial use. Normal wear parts such as hard drives or fans are excluded. This is precisely why the data backup described above is so important: a hard drive may fail, but the data on it must not be lost.

Typical mistakes in online monitoring

We come across the following mistakes again and again. None of them is technically difficult to avoid, but each one costs yield or nerves when it matters.

1. Data only in the controller: the memory overflows and older values are gone, just when they are needed for an expert report.
2. Alarm without a recipient: messages go to a shared mailbox or to a person who has long since moved on to other tasks.
3. Only time-based availability: outages in strong wind remain inconspicuous in the statistics.
4. Curtailments not recorded separately: shutdowns by the direct marketer, grid operator or for species protection appear as faults or disappear entirely.
5. No time zone in the export: at the clock change, values are missing or duplicated.
6. Backup never tested: the server fails and the backup cannot be restored.
7. Display without timestamp: a frozen display is mistaken for normal operation.
8. Access only via the service provider: when you switch providers, the data stays behind.

Short checklist for your existing monitoring

If you can answer yes to these questions, you are well positioned: are all 10-minute values and messages since commissioning stored in a database that you access yourself? Does a fault message at night reach a person who is authorised to act within a few minutes? Do you see both time-based and production-based availability? Are curtailments and shutdowns for external systems kept as a separate state? Has the backup been restored at least once in the past year? Every no is a concrete starting point.

Frequently asked questions

Conclusion

Online monitoring turns a turbine that quietly runs on its own into a turbine you know about at all times. The database server ensures that no operating data is lost. The live display shows the status at a glance. Key figures such as time-based and production-based availability, power curve and yield comparison make visible where yield is being lost. And a well-organised alarm chain ensures that the fault at three o’clock on Sunday morning is not noticed only on Monday.

Manufacturer independence, data sovereignty and a sound backup are not extras, but the foundation. This applies in particular to older turbines, whose data is often not recorded at all today. If you would like to know what this could look like for your turbine or wind farm, please get in touch. We have been active in wind energy for 35 years and have operated our own turbines since 1998.