#127: PD Monitoring │Part 2: The Tech Behind Continuous Defect Tracking

Welcome to Energy Talks, a regular podcast series featuring expert discussions on power system testing data management cyber security and important trends in the power industry. My name is Scott Williams from The Podcast Team at Omicron And I will be your host. Hello everyone In our last episode of this two-part series about partial discharge monitoring We talked about the strategic reasons for implementing a continuous monitoring system To assess the health of electrical assets.

In the second episode, we're diving into the technical side of PD monitoring. Specifically how it separates real PD signals from electrical noise and how to make sure that data you are getting is useful for your operation. It's all about moving raw data to a system your team can trust. Joining me again are Daniel Gephardt and Hendrik Schleifer, who are PD monitoring experts at Omicron.

They will explain how moving to continuous PD monitoring is ultimately about having better visibility into your assets so you can make more confident decisions. they also walk us through SCADA integration. Daniel and Hendrik welcome back to part two of our discussion about online partial discharge monitoring. Hi Scott, thanks for having us!

Hi Scott! Hendrik could you walk us through the setup of a monitoring system? And how it works once that is connected to the PD sensors... to power into a network.

so in the typical set up of a Monitoring device You will have a so-called acquisition unit that you will install on. your Sensors are very close to the sensors. It's actually important that cables coming from acquisition unit go into the sensors, are kept as short as possible because otherwise they will start acting like an antenna and you'll capture signals which aren't related with your machine's partial discharge conditions. So this is step one.

You could imagine it was just a box where you install close to your asset And then depending on the system, if you for example have a temporary monitoring device that would be it. So it's the box that you install and it monitors. but If you want to do fleet monitoring You will typically have multiple acquisition units one for each of the assets That you wanna monitor? Then there is bit infrastructure involved because obviously you want To look at data in one point or get notification if something problematic Is happening.

This would then involve a network connection to each of the acquisition units. And this data is gathered typically in some server installation, where you will gather all your various acquisition units and have a software part that's processing it so that keeps track continuously evaluating the data streams from every device preparing for meaningful visualization and user, because as you can imagine there is just millions of pulses that we're recording in short intervals. It's impossible to look at this.

so we have the machine prepare data into meaningful views. So for example PRPD is a very powerful visualization tool. The phase resolve partial discharge diagram That allows an expert To assess what kind of PD defect were currently looking. Other tools, such as three-part for example allow you to in a three phase asset.

To actually distinguish various PD sources. so we call this clustering. You get the cluster of each individual PD source which allows then in turn to create a peer PD For this particular cluster. and just look at that effect.

And monitoring systems job is I would say The most important part is reliability, so it needs to run twenty-four seven and all of the time. It's not allowed because some hardware hiccups stop its work altogether. So a lot mechanisms need in place for this to run reliably over long periods. Then you have data storage aspect.

as I said before there are millions that need reasonably managed. we need some logic that is processing the data and can keep it for a long period of time. So five to ten, twenty years in which you will still be able look at historical data off your asset And then again You need the alerting aspect. so monitoring system That's just silently on the corner doesn't help.

anybody needs notification mechanism when something critical happens. or let us assume PD source that was looked over certain point different time increased in charge significantly. That would be the point when you raise an alert and that you have to communicate with users. So, either via email or in a SCADA setup How much operator involvement is required after PD monitoring system connected?

You mentioned this happening automatically. Can somebody who installs a monitoring system just leave it until they are notified? We need to distinguish. there's some commissioning part at beginning.

Maybe I'll try an analogy again, measuring partial discharge in an online environment where the asset that you're monitoring is unfortunately never in a vacuum. There are lots of other things around and there's also signals we will pick up at same time as the partial discharge. And this was referred to as external noise right? Exactly!

Noise disturbances or even PD sources from neighboring assets. so assume you have a rotating machine then monitor it transformer connected next to it and the transformers having partial discharge effects, you might see them as well on your end of rotating machine. For the analogy part. so this is...

The problem very similar. if imagine that in a room full people everybody's talking at same time And really interested just one conversation but there all these noise around which makes impossible pick up individual conversations. You would need set-up filter. just reduces the information to this particular one that you're interested in and tries to get rid of all of them surrounding noise.

So, this is into your question - the operator involvement at beginning or during commissioning step - that you will have to configure the asset in a meaningful way so it can actually look like the partial discharge? There's certain risk if configure this correctly. that is similar to a surveillance camera. You point it against the wall, then you won't see anything.

so... That's the operator involvement at the beginning. but as monitoring system running essentially the operator environment should be minimum I would say depends on your level of interest. Of course a monitoring system gives you lot of viable information where can continuously look at development of partial discharge over time.

I think this can serve academic purposes, where you want to look into the development of partial discharge under various load conditions for example. But in a typical use case or if we let's say go back to somebody who is just interested and there are some problems then the idea with PD monitoring system that do not have to interact with it. You wait until you get notified by the system about something problematic. There's an element of trust there.

You're essentially trusting the monitoring system to pick up on a real defect, or developing defect - a potential problem. Tell me in little bit more detail how do PD Monitoring Systems determine whether it is a developing problem and How are operators notified about this situation? It depends upon levels of details that we'll go into but maybe from a bird's eye view source separation essentially, like all of the sources that come in you will need to filter them by origin. What helps us here is a partial discharge defect has fixed physical location right?

It's point of your insulation which gets worse and deteriorates. this where the partial discharges actually occur it's a fixed physical point. The assumption now as signals generated propagates to our acquisition units, they will always take the same path. The frequency response of these pulses as they reach our acquisition unit have a signature that is very unique for this particular point and using this information or assumption about partial discharge helps you build certain tools to sort PD sources when coming in with three-part technique where you have synchronous measurements over three different channels and then you can sort the pulses as the aforementioned clusters, we can just look at individual effects.

You will also pick up noise this way but if you have a mechanism in place that does some pattern recognition on these particular sources keeping track of it nonetheless, but not alert to. On the other hand you will look at a certain cluster that has been identified and if it is partial discharge then you just look over time and then the trend will help you immensely. You would just look at the charge of this particular source overtime And If It increases by a lot versus let's say The known acceptable level before Then an alert Would be issued.

Okay, so given that could you tell us why don't necessarily need to be a partial discharge expert? To make sense of what the PD monitoring system is indicating. I think this depends a lot on how deep do want look into it. in general if The System is set up and commissioned as Hendrik just explained to Us we are quite sure That You get an alarm once something Of importance will happen means of course As soon as you Get an Alarm.

In This Case you might need somebody with additional PD knowledge, not an in-depth expert but some basic PD knowledge to look at that alarm and derive next action steps from it. Is there any immediate action necessary? Do I have to get more information on the operation condition of the asset or things which are not within a system outside this system? All these together...

form the picture so that you should, in an ideal case not need detailed information because you can be sure when the system alarms. this is something meaningful happening on your asset. we should take action of why it develops and changes if there's direct importance to our assets. So basically a good and solid commissioning PDX per knowledge and puts it into the monitoring system in order to just give alarms when its really necessary.

How do continuous online PD monitoring systems catch developing defects during peak loads, when offline tests might miss them? Well I wouldn't directly say offline test missed them...I think that's getting back a little bit. what we discussed at electrical stress on the insulation is just very different during an offline test compared to the online situation.

So it's not that we miss them, they might actually not be happening during our off-line test situations. You can think about your three phases because you have a three phase system and all of these electrical stresses are overlaid in the acid. Well if we do an offline tests normally. So the other two phases are either on a defined potential or grounded, and with this there is not much overlay.

We're stressing our insulation in different ways. so we aren't overseeing defects but actually don't trigger them at all. testing person, they can actually design an offline test to really stress specific points in the insulation to have like a detailed look on that. But nevertheless plant overstressing of one part leads to reduced stressed on the rest.

so you are never in the real online situation and that's the big advantage of monitoring. we are actually looking at the electrical stress the installation sees all the time. One other aspect that is important to a lot of asset operators, Is that such monitoring systems are compatible with their other systems? So we're hearing about industrial internet things.

How do Omicron's monitoring system integrate With modern SCADA Systems for example? To make data more accessible to decision makers We support a variety Of well established protocols. propagate our data to a different system. So, I mean you could for example say can use the iEC-SX one eight fifty protocol to get data from our systems.

or You could be more old school and used the well established modbus tcp protocol. so there is certain range of configurable options that it can use basically what best suits your needs. we would then be able to provide these protocols. like Oftentimes, you can also use a converter or something if it's some protocol that is not supported.

But in the end then we get free range of configuration options. so essentially there are couple of scalar values that monitoring system records periodically and measures every second. That would include most prominently charge value as well as PD rate. with the Omicron monitoring system it's freely configurable.

what you want to get for your SCADA integration part. Once an operator has been alerted, what essentially are they looking at to tell them that maybe there's a problem and they need to take action? Can you describe how this process works?" Well as yourself the operators are alerted.

They look at the system... they find some data.. they found details.... they find PRPDs about the defect,...

they see how it was developing..... And this gives them first indication. I mean obviously if i have a defect which is like very slowly sneaking in and just hit my threshold level now, but I see in the trend that it's been developing over the last three or six months. Then i don't need to stop the machine.

if on the other hand... I do have a defect which just hit and where you see development was in the last two or three days like steep development. then there would be one thing why the operator should think about next actions Additional expert information, be it by the system itself as it might give you a pattern recognition term telling you that this pattern is surface discharge. That this pattern Is slot discharge or By additional experts which evaluate?

That pattern. in this specific case You can then derive next actions. The very good thing they said especially if we talk about machines. They don't die from today to tomorrow by PD.

As We Said It's a smoke indicator, so I should be able to take my actions early enough and do well-based decisions. And not run in fear that my machine will stop working within the next hours." So we're going back to that indication that offline testing itself is not unreliable... in the sense it does not uncover things that a monitoring system would find is simply that you have something continuously looking at the situation to cover that time in between diagnostic tests.

Is that correct? Yeah, that's more or less correct. I mean one example which happened more than once this You are add a machine or add an asset and do it PD measurement there because It has been triggered by whatever reason either By some external influence Or because they did some changes on the Machine whatsoever. And then you realize, you measure some PD in an online situation.

Then the big question comes up and that's also for older machine like was it there before? Because we didn't have any on-line PD measurements or at sometimes We didn't even have PD measurements of the very same asset for a long time Before. so without having that history It is hard to assess if a certain PD no matter of wet level Is critical to an asset. yeah Or If this might have been the situation, this asset has been living in for the last ten to fifteen years without any changes.

And we are again at a point of developing trend. if I do see a trend development that is something where you need take care. and if i see rapid-trend development it's not the absolute level which is counting. so thats why single spot measurements never really giving us good impression on what status of installation?

In both of your opinions, why is continuous PD monitoring becoming a must-have rather than nice to have? Especially for that peace of mind. From this modern day and age we live in data driven world it's just very helpful to know what is going on with the machine. I think PD Monitoring is an essential building block of creating good understanding And I think Daniel mentioned this before.

It's by far not the only indicator of your health, or if you're acid but together with other parameters temperature vibration and rotating machines whatever You have them more of those data points to gather at one point you'll closer you get to build a clear picture off The asset Health Of Your Machine and continuous Data And that is what an online monitoring system will provide you with. It just gives a very clear picture of what was going on, the complete history of development whereas if we do spot checks they're good and helpful.

but If really want to go into direction building this larger asset health view You need continuous data to look at developments over time. I think Daniel also said before The key aspect You just get this gap-free picture of a long period of time where even if things are present, as long they do not get worse you can have good sense on whether it is acceptable or not. And I think that's what the PD online monitoring really helps in. Yeah another aspect we also have fewer people who need to take care more and more assets so workload increases versatile about the tasks they are doing on or around these assets.

This means if I do have a monitoring system which gives me at least piece of mind, one aspect and that is in this case the insulation condition... I can take off my mind and trust and rely on being alerted if there's something important to change. If you could give one piece of advice to an asset manager or even a maintenance engineer, what would it be? Give the PD Monitoring a try.

I think generally is very helpful if you're interested in this topic... to get an overview on your machine at all times and its various developing effects. I'd also say that somebody who's already doing spot measurements during maintenance and gets the first idea of how the installation condition of the machine behaves, I would definitely say advance that picture go to online measurements. And really get that question off your mind and have a monitoring system which is collecting this data and really raising its hand as soon something changes.

it gives you the chance. do more reliable planning yeah? To half an ideal. Is there something coming up or not?

Even if you don't like do the very detailed planning, but. You have an idea about the condition and you might know If There is some extensive Or if you might need to Do more detail. look on The winding heads if you Might want To add An additional spare day for whatever tasks you are Planning. or if You might even say okay I can cut it short.

I'm sure that my PDs and my insulation conditions are good, so i can maybe shrink some of my maintenance tasks. Like even if you don't get away from your scheduled yearly maintenance You do gain some flexibility on how you like Do this maintenance? How you plan it. And also here...

you have the peace of mind That If something unexpected is coming up Maybe On a different asset..that you Can say okay for that machine. I'm sure, i do have a monitoring system on it and honestly very sure that can skip for the next half of year. And push back to take care about whatever unexpected happened.

at another point. Now why should asset managers and maintenance engineers feel confident when using Omicron PD Monitoring Systems to continuously assess condition of their electrical assets? Because we know PD. We had long history about PD measurement with very high precision on different fields as a company and also moving into the monitoring market already quite some time ago.

So I would say partial discharge is something which actually within our blood, it's not that we stepped down developing trend at that point but we can measure a developing trend. Yeah i want to add that we have experience way over fifteen years now, and generally offer a very reliable monitoring system. So something that just continues running twenty-four seven has well advanced. noise suppression techniques gives you the option to look at the individual sources And overall also something maybe in the background.

as Daniel mentioned we have a lot of PD experts. We offer a high quality and the higher level of detail, of PD information even after the alert has occurred to look into data. that gives you just an overall picture in terms of PD what is happening on your asset. Regarding reliability maybe one aspect.

we are just in process decommissioning a PD monitoring system on rotating machine. That's not because this system failed but it was because is now going out of service after almost seventeen years of monitoring continuously, twenty-four seven. So that's just one very recent example of the reliability of our monitoring systems. Daniel and Hendrik thank you both for joining me in sharing your knowledge about online PD Monitoring In this two part series of energy talks.

Thank You Very Much Scott. it has been a great pleasure Hope to be here another time in the future. Definitely Thanks For Having Us Scott. Well Thank You Both And a big thank you to our audience for listening to this and other episodes of Energy Talks.

We always welcome your questions and feedback, please send us an email too. podcast at omicronenergy.com. Omicron has several years of experience in power system testing data management and cybersecurity and offers the matching solution.

Please join us for the next episode of energy talks. Goodbye, for now everyone.