Save Heterogenous Time with airSlate SignNow

Save heterogenous time

now i we go to the real meeting the real reason for this meeting for all these more than 40 people attended today not for my uh awkward introduction so uh it's for the special occasion of having steve adams and colin mcfee they taught couple of courses here in perth so they are also available now they made themselves available to us to do a presentation for us so i think you already know them you i don't need a lot of introduction they publish books which are reference books in saturation height function and in core analysis but i will let them talk more about themselves i hope that they will not be shy about the expertise and the subject is the impact of upscaling on porosity permeability and water saturation modeling in heterogeneous reservoirs so i think steve you are the first one yes that's me the next slide so we need to go to the exit exit and somewhere there must be that one all right let's go there so let's start with the beginning topic of the sub that we're going to talk about is it we're going to talk about heterogeneous reservoirs and the reason we're going to talk about these is because generally we don't do a great job of describing these things in our reserve in our in our work we try but it's really really hard to get all the information and we're going to run through some problems you have with trying to try to quantify what's going on in these reservoirs both and the core acquisition side and also about how do we interpret that information and there's some interesting things i want to show you and i want to get you to think about and i have been doing a lot of discussion about this over we've been working mutually on on a project together over the last few years on a and we'll talk about that case study a bit although we won't identify where it is we've got a very nice data set that's going to help us to show you some some issues and i'm going to guarantee that what we show you are problems that you are encountering and you probably haven't noticed many of all of them you'll know some of them but you won't have seen them all so for those who don't me know me i'm a consultant steve adams i've been around for and then well 31 years now it's upstream oil and gas business um i wrote the shell pretty pressure command you're back in 1993 with another gentleman working at shell since then i've put a new book out available from my website do you go and check that so check that one out on saturation height modeling for reservoir description um so that's a bit of an aside but we've got a turns out over the years i've been made a point of of doing i don't want to do the ordinary stuff i want to do difficult job so i get presented with all these kind of problems to solve which has been great um our business is great in that respect so it's given me a little bit of insight into these things and i've developed sort of a problem-solving event which works pretty well with colin actually when he'll tell you about himself and then we'll get back into them um hello i'm colin mcphee it's been some time since i've been in perth um i'm basically a semi-retired consultant specializing in core analysis geomechanics and petrophysics probably over the last 40 years and i'm the co-author of this book available on amazon.com i made a mistake unlike steve of not self-publishing so believe me you don't make a lot of money from these books i'm not going to retire in the proceeds so that's me i'll be talking about the core analysis side of this particular example so i'll hand back to steve yeah colin's right you don't actually don't make a lot of money even yourself publish either by the way it takes a lot of it's a lot of time and effort to write a book you don't do it for the money you do it because you think you want to pass on the information so so colin's been really really good to donate us his knowledge to the industry for not very much money buy on amazon.com so what are we going to talk about upscaling it's a word that we encounter a lot in our business so we're going to talk about it here in a little more detail and probably a lot more detail than you're used to seeing and we're going to talk about why it matters and then we're going to talk about problems with the core sampling and how that might impact on what we do with our data and what assumptions we are making and then we're going to do some simulation exercises just looking at homogeneous reservoirs and upscaling in there and then we can do some upscaling and a simple binary binary you know poor reservoir good reservoir system to show what happens when you do it there in a model then we're going to look at a case such study and see what happens so we're going to look at some very high resolution data one centimeter type scale porosities and permeabilities and then we're going to do look at what that looks like on the logs look at some core plug scale data and some then there's some wireline log scale data we're going to see the differences between these data sets as we upscale between each different set and it's quite interesting how things change and there are some assumptions we're making we'll discuss these as we go and at the end we're going to do some talking about comparison against well tests see how the performance is in this particular case study and i guess the point i'm trying to get to across is that you know permeability will change when we upscale that's because the relationship from processing permeability will change and that's fairly well understood although not although not we don't always do anything about it and i'm going to show you that we should but we don't recognize this is actually an impact on saturation height modeling as well unless there's a no if there's a permeability permeability relationship between with water saturation and there usually is so we'll talk about at the end we'll talk about some possible ways to deal with this and whether you should do anything about it and the answer of course is you should do something about it and we'll take it from there so upscaling what is it it's when we basically take measurements at one vertical scale and what we're trying to do is we take them on a call plug scale which is about four centimeters round about because we're looking at horizontal plugs and we've got log scales 15 centimeters then we've got a static model scales usually about a meter you might be lucky if you're working at aramco well one could debate whether that was lucky or not and that they don't upscale at all they absolutely they work their models nowadays on on a log log increments so 15 centimeters and that's both the static and dynamic model so some of these issues just appear if you have enough computing power but we most of us don't so we have to deal with this so what actually happens when you're upscaling you want to be able to preserve the performance behavior so you want to actually make sure that when you go from a core plug scale to a log scale to your static model scale that the reservoir has the same volume and performance characteristics across the appropriate scales and it's not as straightforward as you might think so why are we concerned well first of all we acquire core plugs at one scale which i said about four centimeters and we use those to construct relationships that's our porosity versus permeability and a saturation height function then we apply those relationship with a wire line log scale so that's a 15 centimeter scale so are we making a mistake there that's the first question then we log scale the log scale data average up to your static model and which will be the one to two meter scale depending on the dynamic or static model what we would like is the reserve performance and the reason why description at each of those scales should match the actual performance of the reservoir if we actually produced it at that scale that's what we really want so what's the problem well it turns out in a homogeneous reservoir there's no problem so the homogeneous reservoir we've got no problem at all and heterogeneous problem revised we do have a problem because porosity which is volume volume type measurement upscales just fine but permeability is not a volume measurement penalty is a directional measurement and it turns out that doesn't upscale just fine and we'll show what then what i mean in a minute and water saturations because they're usually related to permeability don't upscale properly either and that's something we haven't really addressed previously that has been addressed before in the literature and i'll talk about that in a minute but what we also have is problems with the sampling so how do we make sure we have samples that we can actually use and that's what i want to leave colin to talk with in a minute but we'll get to that second second so what we're going to do we're going to start off with discussing the problems with the core and on which you're in the core the reason says reason this is important because it is core is what we base our primary relationships on we base our processes and our perm abilities and the water saturation height models on and they're going to say well to do that then we'll talk so colin's going to talk about the core problems then we'll talk about some simulations which i'll show you some simulation results and then we'll show what happens in the binary system and then we're going to illustrate what happens in a real real core case and i've got some really interesting core and core plugs to look at as you'll see so colin's going to talk about routine core analysis problems and unfortunately it's not as nice as we'd like it to be yeah this is um example is a real example of some of the challenges you face not only in upscaling but actually getting measurements this is a fairly heterolytic sandstone reservoir um i'm going to talk about the problems we had actually going to get plugs that you could measure things on i'll talk about how we had to change the system and then i'll talk about how some inaccuracies in the measurement can feed into the the problems we had and you'll see that with what steve is talking about in some of the upscaling okay um this is just one well 400 plugs one and a half inch diameter plugs taken at roughly 1 30 centimeters 25 centimeters plug the kerosene um the clay content in this reservoir is variable um initial xrd suggests it's between 10 to 16 percent um we wanted to condition the core for total velocity measurements so in that case we would try to do hot socks like cleaning and conventional oven drying at 106 degrees celsius um to make sure that if there's any clay bound water this could be removed so that we could be measuring total porosity and when we started cleaning the plugs we've got a very high failure rate uh but in particularly in the heterolytic intervals where we had sand shale laminar sand shale clasts and what we realized probably was that hot toluene takes the water out of the clay and we think this boundary between the sand and shale was dewatered by the hot toluene cleaning and that promoted failure along the bedding planes of the of the core plugs so we looked at lower temperature cleaning and for those of you know this you could go to cool soxlet but this that that would take years in this case um we looked at flash cleaning but it's 650 a plug found it was a little bit expensive when you get 400 plugs so we use this total immersion soxhlet apparatus the advantages of this is its lower temperature that was a principal advantage rather than retention of any clay bound water and reducing interfacial tension the whole thing about this was it was lower temperature so the aims of the subsequent measurements were to try and minimize the risk of plug failure and capture as much polypem data before the plugs fractured so it was a bit of a weird routine core analysis program basically involving three cycles and part of this was driven by the fact is that permeability is only possible on an intact plug you've got to have a cylindrical plug on fractured plugs on my shaping plugs you can measure porosity and grain density so what we're looking at is determining ambient condition porosity and clink above permeability in this particular well at reservoir fining stress so the first cycle was basically we cleaned it in toluene and humidity oven dried the samples we measured porosity permeability and grain density on every sample this meant that there's a risk of salt or residual oil in some of the plugs so we'd expect perhaps the porosity to be reduced grain density to be reduced and the permeability to be reduced so after this we then cleaned it again um in the hot stock in the the the total immersion soxlet with methanol and then humidity oven dried it at standard conditions this runs the risk that in australia sand any clay by water might still be retained and the the porosity is somewhere between total and effective and then in the third cycle we basically did conventional oven drying at 100 degrees celsius and that should remove any cleveland water um this technique and in retrospect we maybe could have looked at other other um ways of protecting the plugs as well but generally the far fewer plug failures and this was then transported to the other three cord wells and what we found is if we look at the um this is the um the change in um porosity and this is a change in permeability between um between after toluene cleaning and basically we found it was little significant increase in the corrosion permeability another was tally on its own probably was quite effective at cleaning what we're probably more interested in is the change in porosity and the change in permeability after conventional oven drying after cleaning and humidity oven drying so cycle three to cycle two so that's a difference in porosity that's a difference in permeability and what we clearly see here is in the the the kind of lower permeability samples and the red are fractured pipe samples and we ignored those um but when you look at the intact samples which are shown as the large dots what you see is as you would expect a small increase in porosity um in those samples but generally when you get above about 16 17 percent porosity there's hardly any significant change what you're seeing there is just random error and that's within the acceptable error range for porosity measurements so we couldn't see much there same with permeability apart from in the 0.1 millidasset and below this is nothing to do it actually masked any effects of of in enhanced drying and this is was due to basically inaccuracies in the measurement there was probably around about 60 percent error in plug samples with uh permeability less than 0.1 of a millidacy so in terms of measurements we just applied a cut off and basically handed over the data in terms of a quality flag so is green amber and red amber is basically um porosity data and samples which may have fractured green was the best data set that's the one that we've got an intact plug and porosity permeability and grain density have been measured and have been quality controlled okay now before i show you what the data that colin's tolerance talking about how it all came together what it looked like i'm going to show you some simulation exercises we ran and the reason i'm going to do that is because the simulation exercise will run on homogeneous reservoir just to illustrate that the methodology we work we're using does make some sense in a homogeneous reservoir then we'll run it on a simple heterogeneous reservoir and then i'll show you how the core looks and what the core plugs look like for this particular case and the reason i want to do it this way is because you can you can get an understanding of how you might expect things to vary and then you'll see how it actually works and see how things go so well this is not it comes out much better on the screen over there than it does on here now i just got to turn this around so we can see so what we have here is a straight up homogeneous reservoir 50 milli darcy rock 20 porosity and um what you need to do is look at these curves on the right hand side and see what they are so this track is a peripherasy track showing one thousand to eleven hundred million dollars from meters sub c and it's just homogeneous this is just a model permeability is on this track and see around fifty milliliters and on the right hand side is a water saturation track now what you need to know here is that water saturation is generated using a j function the saturation height function so it's using porosity permeability and a j function so we've got a transition zone at the bottom you'll see two lines on these curves you see a blue one and an orange one the blue line is always the one centimeter resolution curve so i'm going to show you a series of displays so the blue line's always one centimeter and the orange one is the upscaled version so in this case it's upscaled to five meters and you can see that there's absolutely no difference between the porosity and permeability upscaled no surprise it's a homogeneous reservoir there shouldn't be you do see a small difference at the bottom of this earth hang on turn that off small difference in the water saturation to the bottom and that's just down to averaging over a transition zone and the layer thickness is just too thick so what i'm going to tell you now and you have to take my word for on this one is this is a 5 meter sample averaging and the four centimeter the 15 centimeter one meter and two meter they look almost exactly the same only just at the bottom end you'll see they fit better to this curve at the bottom so that's a homogeneous reservoir let's look at this heterogeneous reservoir so this is what the reservoir looks like we've got five meter beds and it's point zero two milliliters on the on the on the on the military here and on the other side it's two hundred milliliters we've got a linear porosity relationship so if i have to draw a peripheral line it would join those two dots so that's the poro prime relationship on the bottom right you're just looking at two permeabilities so it's either 0.02 millidashes or 200 milliliters that's the first point you need to get across so that's it one centimeter so let's see what happens when we take these and we go up to a log scale so log scale data is 15 centimeters odd and you can see straight away even with five millimeter centimeter bits the model changes so our purified model is changed from the straight line down here to this orange line and that's because we're upscaling the permeability in a linear linear fashion and displaying it on a log plot which is fine it's behaving as you would expect to do if you investigate it so that's that's what we expect to see what i'd also like you to note and if i actually do the numbers and calculate even though you can't see much difference between the orange and the blue on these curves there is a three percent difference in net times porosity times hydrocarbon saturation already so the debts decreasing the ehc is decreasing by three percent just going from the core so one centimeter to a log scale so it's already decreased there just clear that one in mind we go to one meter scale on a static model so you can see over here these points are still on the same line but their location on that line is changing as we go and i can see more differences on the the orange and the blue over and the permeability and also on the hydrocarbon saturation track what's um interesting here is we're seeing this impact even only they're thick five meter bits this is still a change it's just a way that it's all about how um how we do our boundaries these days so when i first started in the oil industry we first didn't use compute we used the draw maps by hand of course and then we go the first static models we used to build the boundaries always used to be at the deformation boundary so if there was a change in reservoir properties we would put the formation boundary with that location in our static model these days we just do it on a depth basis you know every one meter or half a meter is our vertical thickness thinking about covering the geological boundary so maybe there are sometimes we don't always improve things changes should really happen at the geological boundaries if you can do that anyway so digress a little bit so something's changing there program going up to two meters and again seeing the poor open model changing so we're seeing this over here so that's pretty much as we expect what you do now is notice if i go at the water hydrocarbon saturation so you look at the the orange curve now that's the upscaled version of the data and you can see yeah we're starting to get some issues in our water saturation and our permeability models um this particular case it gets an increase in net times priority times hydrocarbon saturation of 3 over the log scale data which in turn was already oh sorry go there was uh one log scale where's the log scale there it's three percent down from the score data so that's no that's okay we can go to one up to one meter that's okay um two meters okay but when i go to a five meter scale this is why you don't use thick beds by the way so the poor repair model has now got either everything up with the high stuff and sorry everything down at this low low perms but that has a significant impact on your ehc so it's increase of 27 over your log scale data so the saturation height model is now wrong and unacceptably wrong very upscaling if you don't adjust your perms so let's have a look and say what is what happens in a real case so i'm going to show you some very high resolution data from a core i'm going to look at core plugs we're going to show you some logs and then i'm going to show you some upscaled static models and the way i'm going to do this is i'm going to show you the raw data first then i'm going to show you what happens when you actually physically upscale the data so here's raw data on the right hand side you can see a lovely piece of core from 2619 through to 2626 and i couldn't put all the core on here otherwise you wouldn't be able to see it but you can clearly see this is not a thin bedded system it's a heterolytic we haven't got we've got a lot of different things going on in here so you can see what the core plugs are um shown on the core and we've got a missing section through here it's preserved sorry so we've got a whole lot of stuff going this is obviously the best reservoir element but you can see oil staining in these elements through here through here and essentially there's nowhere here you could put a core plug and guarantee that it was going to be perfectly homogeneous on the left hand side you'll see a couple of logs one says phi high and k high and they're high resolution prices and permeabilities how they how they calculated them was as a paper in itself but essentially it's based on a uv trace through the core and a transform derived from the homogeneous core plugs that we did get so that's a quick summary but we have a high resolution version of the price imperability over here we're going to use those in a minute and this they don't look i mean this for instance if we look at this unit here it doesn't look as though it's changing a lot that's only a scale issue if i was to zoom in on that you'd see a lot more changes taking place that's just a display issue okay so what are the core plugs look like so we're going from so let's just get this straight we're at a high resolution here one centimeter the next slide is a core plug scale so what i've got on the right hand side i've got my high resolution porosity and permeability and at the core plug scale i've got the rear points the core plugs porosities and permeability and there's a blue line is the mini permiameter and a mini permeability is effectively a core plug scale type measurement rather than except there's a whole lot of core plugs next to each other it's not not the same as a really high resolution management measurement colin might argue a little bit but not too much what you need to do is also look at the core plugs and the core plugs look much clearer on the screen than they do on this screen by the way so you can actually start to see a bit of texture see these are supposedly homogeneous core plugs they're clearly not we've got these features in here you've got cracks running through here you've got another crack running through here so you can see why there were problems measuring the reliable process and permeabilities that shouldn't really be a surprise so other things that might be at a core plug scale would be things like an fmi or an ultrasonic measurement and maybe the dip meter resistivities you could use to give you a core plug scale type measurement and here's some more core plugs just to show you that um you know even these this one here is probably the most homogeneous plug that we've got a photograph of and even here if you look carefully you'll see we've got some cemented streaks and through here so we do the best you can and that's the nature of our business we do the best we can but sometimes we have to put up with these kind of these kind of problems okay now we're log scale data so what i've got here is again over this porosity track you'll see the high for high resolution porosity a permeability mini permeameter is on there now i've got a log porosity and you've got the wireline log curves on the right so there's a log porosity on there and there's a log derived using the corescale poro premium model that's the light blue over the radon all of it so the water already you're seeing over here you're seeing that the high resolution data is well you go down to the mini permeator it's being sucked down a bit and then we see again when we look at the log scale data it's no longer representing the peaks that means if we use this description we're not going to describe the dynamic performance properly okay you might get the volume performance okay volume okay because the porosity average is probably all right but you will not get the dynamic behavior of the reservoir correct now if i upscale that again to a one meter vertical scale like a static model that's the magenta lines you'll see on here so you can see again shower fraction porosity but we're so we're starting to lose really are losing porosity uh resolution and certainly our permeabilities are definitely being well and truly impacted by that upscaling activity and you can see of course this is where i was referring to before on the hydrocarbon saturation track you'll see because we haven't got the bed boundaries at the same location as geological boundaries we have got some mismatching issues that's standard these days unfortunately so i'm not going to go to the static model scale because it doesn't serve any more purpose to illustrate than we've seen already so there's a few points you can get out of this straight away and that's a first of all that you process the way we work in our business is we take our porosity permeability transforms derive the core plug scale and we apply them at the resolution of the logs and we assume that's a real right thing correct thing to do and we'll see in a minute whether it is or not when logged derived and let me oh update your virus protection hit the cross okay oh sorry i'll get it timing's everything isn't it when okay the other thing is if when you when we log to our log derived processing permanently upscaled to the poor resolution center of the scale of the static model that averaging will change the relationship between porosity and permeability so we need to be sure that we don't use a log-scale puroprim model at a static model scale okay so if we're going to put porosity and permeability in the static model we have to take the log-scale perm and upscale those with the porosity okay um so yeah that's basically down down to the the difference between volumetric up scaling and directional which you use for the terms um a whole lot of stuff on there on that but at the end in the end we actually have to verify upscale permits against well test to see whether actually reasonable not it's the best way to confirm the success of any upscaling so i'm going to go back now to the high resolution data so i showed you those high very high resolution porosities and permeabilities before that's these tracks so these are exactly the same displays i showed before back here same type of display only i'm putting real data in there so this is real data and the important thing to note here is this interval from 2622 to 35 is the interval that was well tested so we actually do have test information across this interval so at the moment that's my log poor operating relationship don't support permutation just a straight line okay that's at a one centimeter scale so i'm going to up scale that now hang on wrong way i'm going to upscale that now first thing that's just going to a core plug scale so from one centimeters to four centimeters look what happens to the porosity to permeability relationship so already it's gone this is an important part note to point to note we don't go the permeability doesn't decrease it doesn't go down it goes up it's an upscaling is it so it pushes a whole lot of stuff up at this end that's just going to the core plug scale and we go to the log scale let's go a little bit up again and the reason this happens let's just go back to here for a second sorry i'm jumping around a bit so why does this happen well it's pretty straightforward if i look at if i look at this relationship here what's actually happening is we've done that averaging with the different beds but if i start from a different place so the bottom end of these points is a different place on this curve then we get a whole family of these type of curves going up so that's what we get in real data it's a whole family of curves going up so expect to see these kind of things so what we see in our upscaling operation here is exactly what you would expect if you think about it the problem is we haven't been thinking about it so there's a full that's that's your up to four cms scale up to a log scale and obviously we're using losing vertical resolution which is why we have less points in these so so if you're going to draw straight lines draw a line through here it would no longer be straight your best fitted model is no longer through this through tonka straight through here and you can start to see we're losing a little bit of vertical resolution but not so bad maybe the water saturation is being lost a bit here and i'll give you some numbers on that in a second but when i go to a static model scale if we don't model these things properly we're going to lose we're losing a lot of a lot of the performance behavior so i mean look at this water saturation averages so there's a whole lot of things going going off here we have to be really careful so poor open we might get right the water saturation actually has an impact and we'll talk about that in a minute dynamic model scale then there's just a whole lot of stuff going wrong but you see that what i want you to notice the poor open model has changed so we go from core photograph scale to a core plug scale to a log scale to a static model scale so you can't use the same poro prem model for different scales that's the first that's a really important point i want to get across but there's some interesting things going on here you think well so what do we what can we do what is what is this telling us well this diagram is trying to show you something a little bit more complicated what i'm showing here is core plugs in light blue on the right overlaid is the log derived log scale upscale so i'm taking let's start again i'm starting with the really high scale data i'm upscaling it to the core plug scale to give the light blue then i'm upscaling that light blue to the orange data which is the y line log scale data then overlaid on that in green is the core plug data that's porosity and permeability from the core plugs and what's really interesting about this is first of all that orange and the green the green pluck or plug data appears to overlay the log scale data pretty well apart from this point at the bottom which you may have noticed is below 0.01 milli darcy which we can tell from what colin said before is in fact a load of rubbish those measurements are unreliable so we can discount that stuff but what's interesting is this actually confirms that we can use our core plug data and assume it's valid for the y line log scale okay so this is actually showing that we can do that um the bottom if i make this assumption the assumption that we can do that is valid in this case i'm not saying it's always valid but it is valid in this case i haven't run enough examples of this to see if it's valid everywhere we are in our business we assume that it is but it's probably something we should check occasionally okay again so probably from the proceeding it's clear that we shouldn't populate statical dynamic models with perms based on porosities in those models and the log scale prosthetic transform and the reason i mentioned that is because i'm still seeing people do that okay don't do it it's going to underestimate your reservoir performance which is one of the things you really don't want to do they should upscale them together upscale them together and say which upscaling process you depends on your geology and fluids but most cases especially in a heterogeneous reservoir most cases are we're assuming we're getting lateral flow because the kv goes to zero in these things vertical primarily goes to zero so what can we do well we can test this against some core some floaters so we had an interval 26 20 26 35 was flow tested giving a kh around 1588 milli darcy meters now i want to compare that performance with what i've estimated from my upscaling of the logs so the high resolution data upscaling it properly but you've got to remember that the decay i'm getting from my logs is a single phase permeability and i flowed oil in the presence of water as it so the weld test is a two-phase permeability which means if i'm going to compare them i have to transform my upscaled core date upscale we'll call it core data upscaled high resolution data i have to transform that to two sec two phase data and i've shown that down in here and there's a table with it all but essentially it summarizes after fluid correction so that's uh correcting for two phases kh from the high resolution data is 1843 meters which is more than we got on the well test while you're conventional or poor opera model so that's one you would have got from a log scale data if you just did your normal log scale data as a thousand or 64 which is underestimating and now remember we're just doing simple linear averaging so if you had a harmonic or geometric mapping averaging it would be worse okay so what we're seeing here is that if we don't do this we're going to underestimate our dynamic performance and the reason this i said the highways of data overestimates the kh and there's a real reason it's good simple reason for that and that's because if i go back to the core photograph that's because the beds in this aren't linear they're not straight thin bedded systems if it was a thin bedded system then the chances are the perm as we've described it in the world test would be much closer together so what we've got here is a system where the oil has to flow can't flow in a straight line horizontally it's actually got to do a few wiggles on its way so we've got somewhere between the conventional and the high resolution approach and that's the 1588 we saw in the world test so in reality in reality the assumption of horizontal flow by using the upscaling high resolution data is only partially valid in these kind of rocks so if a thin bedded system it might be much more valid but then this this approach would work there too by the way okay now this is a bit i haven't sort of touched upon a little bit before but water saturation is also there and also impacted this because your saturation height functions unless you're really lucky are permeability dependent and permeability again is scale dependent so in this particular case i did keep track of the numbers so when you do the upscaling and it's a very interesting bit about this the saturation height function will generally over rest over estimate oil saturations and if we're using a log scale data as your baseline it's a 14 overestimation unless you correct your saturation height function uh ellen johnson put a paper in the sbwa last year in london talking about all this stuff in here does he made a suggestion there about and ellen lives in aberdeen that's joe collin knows him well um and i had a good discussion with joe allen about this at the time um yes helen suggested you he's a bunch of people to use a second perm ability in your model that you can use there to describe the saturation height function which is different from the dynamic behavior now i'm just a simple simple petrophysicist that seems to be an overly complicated way of doing it and confusing way to do it because people will go well which primarily do i use for what so it turns out it's much easier just to put a scalar into your saturation height function than to match a log so the way you do that is you upscale your poor opium data then you run your saturation height function and then you tweak the saturation height function so it matches the correct volumes where you need them so you do actually have to test it you have to test it in some some example what's the best way some representative wells from your field so you really do have to test it and check it it's interesting because when alan pointed this out it rang a bell in my head and i went back through my files that's the other advantage of being a consultant by the way is that your files go back a very long time because you don't lose them every time you change company and i found some far sampson i'd actually looked at this exactly this problem back in 2002 and i'd got sidetracked off it that uh you know i'd noticed exactly the same thing so this is this is a valid concern and i've checked a bit we've done some work on this so check your saturation height functions okay so here's the summary on the water saturation no impact in homogeneous reservoirs you don't have to worry about it it only has an impact when you've got heterogeneous formations and it turns out it only really has an impact if your permeability is less than about 50 millidashes so if you've got a good reservoir most of the reservoir about 50 milligrams you probably don't have to worry about this issue but if you have beds in your reservoir that have got hydrocarbons in them that are less than 50 milliliters you should check okay and you should also check if you're getting the more closer you get to the transition zone the more it becomes an issue depends on thickness of your beds just one of those things you need to run through the exercise and see so what else we got um yeah so at the bottom of the important point is you need to carry out some upscaling trials don't trust that the assumptions that you've been making are always valid because we found in this case that some of them are some of them aren't so the fact that the porosity and permeability uh from the core plugs are valid for the log scale um the reason that's worked is yeah it's this is one of those nasty nasty things you don't want to say but i'm going to have to say anyway the reason that actually worked is because what what actually happens one of the problems we have and when people take core plugs is in that we want as a petrophysis i want those core plugs to be regularly sampled so i'd like them every 30 centimeters so that means they're really easy to run a filter over and compare with the wire line logs but what happens in the lab is the lab technician goes along and you'll go to that 30 centimeter point and you'll go well that's a nice homogeneous bit of rock i'll take the plug here then he goes to the next one and he goes oh if i take a plug here i'm not going to be able to get a measurement the plug's going to fail so i shift it a bit to make it a more homogeneous measurement measurement so and he does that as he goes through the core and if you've seen that photograph of the core here he's doing that a lot but what that actually means is i'm getting biased sampling okay so ideally we don't want to see that but in our case it's worked because it's actually made what it's done is it's given us a baseline to get a poro per model that's mostly homogeneous plugs so when we do the upscaling we can actually see what goes on but you have to be aware that that's what's happening in order for this to be useful if that makes i'm not sure if i got that clearly through or not but we'll see when you ask questions so what are we going to say to the end we've got modeled and real data and we've used them to illustrate some of the difficulties in upscaling heterogeneous reservoirs and i want to make this point this is one case which we've got lots of data for it's by no means the only case where i've seen this as a problem or an issue in fact i would say probably this is an issue in most reservoirs yeah if you've got a really big thick reservoirs and off in the northwest shelf the big thick stuff probably not going to worry about it but you've got the thinner stuff you probably do need to worry about it the other point all the measurements we have we can acquire have a vertical resolution that differs from the measurement type so what we really want is what we can't measure what we really want is the artist that's what i'm trying to say there and that vertical resolution does impact on the relationship between the volumetric properties so that's porosity and water saturation and permeability so the vertical scale you're working at has an impact on those issues on those relationships so that's porosity dependability and saturation height so only in homogeneous reservoirs can you assume that those relationships between pricing and permeability or saturation do not change with vertical scale only in homogeneous reservoirs otherwise they do so i'd recommend you always investigate the impact of that upscaling on the relationships you are using they're not hard to do the the hardest part of that this exercise we've been through here isn't the upscaling exercise it's actually just getting the data so that's the hardest part so just think about scale and i'm going to leave you at the end with this wonderful photograph of a core that you said on a set of logs that doesn't look anywhere near as bad as that but it is that's the real world so with that um we'll take some questions [Applause] there must be questions yeah well just sit close to you colin use this steve thanks for that interesting talk those who don't know me i'm trevor mcgee um steve and i go back at least 25 years i think it's 27 when i recalculate it but steve you mentioned uh the regular sampling and you know going for every 30 centimeters if possible if that isn't possible do you think there's a case to recommend sampling every facet what i'm yeah probably one we can both answer but the answer would be yes i would recommend what i'm after is a set of rep as a representative set of core if i'm not sampling regularly it means i makes it harder for me to compare with my compare to upscale my core data to the logs but i'd rather have a all the porosities and permeabilities and fashions sampled than not have them all there do you want anything anything to that yeah i mean you know sampling's a real issue one of the things we didn't mention on this is um in one of the wells we used a scratch tester which is a continuous strength measurement but strength is inversely related to porosity and to permeability and that was very revealing when we actually sampled on the basis of that so we looked back and look at the heterogeneity and classified the strength heterogeneity in fact we chose a lot of scale samples from that and quite successfully whereas in other worlds sometimes the you're right you you've got some you know the the spacing isn't random but it's you know it's biased generally i guess that i'd like to say is that if you've got you the high resolution data i showed which came from the core if you've got that kind of approach and you've got the core plugs and you can upscale the two together it doesn't really matter where they're taken from they will tell you whether you've got the fashions correct so yeah i'd rather have all the flashes than a rig than regular sampling ideally you want both but if there's a choice that's what i make thank you c for the interesting presentation i have two questions basically the the first question about like the term heterogeneity as you mentioned because heterogeneity is like a qualitative measurement did you like uh think about a way to quantify heterogeneity like how heterogeneous or how homogeneous and then we can put a cut off that okay this reservoir is homogeneous or this reservoir is heterogeneous and then apply the relevant corrections and so on this is the first question the second question is about like the differentiation between anisotropy and heterogeneity because there are two different things and we know that permeability for example can be anisotropic rather than heterogeneous in the reservoir and okay as you mentioned always in your presentation that it should be taken into account like sometimes separately so there are questions if you can answer quickly thank you very much okay well let's um can you remind me of the first question first well please yeah the first question was about quantifying the heterogeneity okay yeah right let's say quantifying heterogeneity what i would say i mean it's easy to say just quantify it everywhere but that's not really going to help you so what i my view on this is is basically you look at the log scale data if it's varying at around the log scale or less then i would term it heterogeneous okay so if it's so if it's a 15 centimeters or let's call it 30 centimeters so 2 2 increments it's varying below that scale then you're going to effectively be heterogeneous if it's varying at a scale that's greater than that you probably don't need to worry about it too much but you know having said that it's it's something i think you should check okay so that's that's the first thing um you agree with that yep and the second question was on anisotropy right but you're absolutely right permeability is one of those things that we make a lot of assumptions about and what we're going to do what with the assumption we're making when we do this exercise is that first of all that the kv is is not very large and generally for these type of rather wise it's not so if i upscale any of these beds that you see on that screen up there over about two centimeters the kv goes to zero so so that's essentially gone so we're now we're talking about the two horizontal permeabilities so what we're effectively need to know is if we knew the direction that the things were deposited and we can actually back out a the two vert horizontal perimeters maybe we could do something on those lines but as it turns out we don't usually measure that so we're making in this exercise as we've run here we've made an assumption that the k h's are all the same which yeah it's probably yeah whether it's a valid assumption or not i suspect it's probably not um but in some systems it would be but very effective hopefully what we're doing is because we do see a reasonable relationship between porosity and permeability on the more homogeneous plugs so we're assuming we're seeing the maximum or maximum horizontal permeability is what i hope we're getting rather than a minimum and if we do that then we're probably not going too far off as far as performance goes in shallow yeah my name is musa from corelab and uh yeah and uh i just want to make a point about sidewall cores if you just went for sidewall cores and no core how would you uh determine your heterogeneity um what can i say i'm afraid the industry is moving to save world core and digital core analysis because um it's essentially cheaper we've not been able to do this with the sidewalk not at all in fact a lot of the sidewalks failed didn't yes they didn't yeah so we would have even less data and there's nothing beats a full diameter and the question about an isotropy of course if you do hold core analysis you can also get an idea of the the anisotropy in the horizontal direction this wasn't particularly necessary here but in part of the cabinets and we're doing a lot of work in the middle east um that becomes very important whole core analysis which you can't do with um sidewalk you have to make sure that's core is orientated if you're going to do the whole whole core analysis of course uh one more stephen i hope i don't dig myself into a hole here trying to explain this but i think we had an example where uh the operator the been describing the reservoir exactly as you had said and um effectively overestimating water saturation when we looked and we went to the the core we we find this was a laminated section so our approach was actually to look at a plug and let's say the plug was for instance 50 shale 50 sand then we said well the shale porosity is zero the shale um permeability is zero assumption but probably not too bad and then effectively doubled the sandstone porosity and and applied a net to growth to the to the reservoir and we find we got something that we we felt we got a better history match with uh is is that approach you would uh recommend i understand where you're coming from with this it's um and you're absolutely right in that there's no no oil found in the shale shale of the elements um usually if you're using um it depends whether what sort of how you're doing the work but it's generally the rule of thumb of doing a doubling is probably not a rule of thumb there is some change you certainly can work out what the correction should be and that's essentially what you do in a thin bed type approach so yeah it's there nothing wrong with that approach per say but you just have to remember to discount the volumes by the show you've removed the only point i would make about that though is you've got to be aware that as we commented before is that core plugs are selected to be homogeneous so when they go through to drilling the core plugs they are already being selected there's a selection process so you you may be if you just use the call plugs alone you're going to over overestimate what's actually happening you need to actually look at the whole core to see what's really going on thank you thank you i'm a drilling guy so i look at some of these things a little differently i was wondering if we drilled a well vertically and completed it and selectively perforated it based on the mud log and rop and gas and then we have a field there so then we do three multilaterals from that vertical and say out of 1000 meters in three directions and then again selectively perforated it and sand or acid cracked the sand to say get a 200 meter diameter on your permeability to the perforation that you selectively perforated and then produ and then multi complete it leaving the vertical separate and then having your three commingled multilaterals would you be able to calculate how you get a greater 40 percent greater return on asset with that method versus what we can recently use i don't think this is really the right form for that question all i can tell you at the moment is if that approach wouldn't have worked on this of this particular reservoir because it's not laid out that way it's not it hasn't been deposited in that system so um yeah it's it's uh that's what's the economics the economics change the numbers so yeah we're not really in a position to answer that question so it's really beyond the beyond the scope of our um expertise in this particular field any other questions yeah i'll go steve yeah uh chris woods for those that don't know me um thanks for a great presentation guys um i was just interested in um going from the one centimeter scale which seems to be a linear and then you show the actual core plug data and it's quite high and it looks similar to the stuff that it ends up being when you upscale so can you maybe talk us through why you start off with a linear thing which doesn't really represent the core plug data and then well how different really is um the actual poor perm cloud that we normally use to that sort of general one meter scale i know somebody spotted the question i should have known it was you [Music] that you could spend all day describing i could spend all day describing how we got that but essentially what it comes down to is that poor per model we started with a high resolution data is based on the cleanest core plugs that we got so just taking those on like those alone and leaving the ones that happen with it that had any any signs at all the shale are out so it's just a really clean stuff that we were short and the measurements we're sure of so that's essentially where that comes from it's it's a it's a simplified model yeah yeah it's it's um so what's actually happened is that yeah this is a maybe we should take this one offline because it could go on for a while but essentially we've calibrated that's back to the back to the original core so we've got the high resolution data we've calibrated specific points that we've upscaled to that core plug data and bring it back to so that the end points are tied it's the bit in the middle that's not tied so yeah there's a whole lot we can go on about but i'll maybe we talk about this one offline any more questions one last question looks like no more questions so uh we have two presents two tokens some medals coins from wea we have a symbol of professors and spwla i think they are gold coded but it's not enough to retire on them either so thank you for doing this presentation for free how to say and thank you thank you