History Match Complex Model
One of the best approach to deal with complex and large dynamic model with objective to improve its history match quality is by break it down into multi bodies to asses each part separately under one structure. With the available commercial assisting history match tools and powerful high performance computers, it would be so practical to load as many defined variables in the first attempt to evaluate their impacts on history match quality with objective to keep only the top 10 ranked variables that identified by tornado analysis plot for further adjustment through additional runs.
In the following example, over 3000 variables were designed to evaluated their impacts, which basically it would require thousand of runs targeting the following items:
(A) To change vertical i.e. (PERMZ) and horizontal i.e. (PERMX&Y) permeabilities multiplier per each dynamic model layer by 10^X where -1.0>X>1.0
(B) To change Kr's curves end values, corey expansion and Sor's ranges to accelerate and slow down water & gas movement.
(C) To change faults corridors grid cells permeabilities in order to accelerate water & gas breakthroughs parallel to faults as needed to match the observed measure data
(D) To change wells regions grid cells permeabilities individual within small range to avoid creating localized adjustment surround wells.
(E) To change globle permeabilities multipliers with very small range.
(F) To generate Sw_log uncertainty profiles by create High case (Sw_log^1.25) and Low case (Sw_log^0.75) where (if Sw_case<min Sw_log then Sw_case=min Sw_log), then to use Fast-Track approach to create new SATNUM input parameter in order to be loaded into dynamic model without need to change saturation tables.
(G) To change faults transmissibility for pressure and flow match.
(H) To change OWC's and GOC's depth interval.
Porosity Model
Permeability Model
Faults Corridors
Well Regions
Height above FWL depth parameter
Static model water saturation model (Sw_log)
Dynamic model water saturation model (Sw_Pc)
Rock Type scheme (SATNUM)
Researcher Reservoir Simulation
3 å¹´A very interesting article about how we can explore the use of complex models?to improve its history matching, Thanks for sharing Faisal.
Principal Reservoir Engineer (Simulation)
4 å¹´@All, one of the key messages I want to highlight to you that try to avoid using global adjustment like using KV/Kh single value for the whole model where you might improve your model history match in one side while destroying your model history match on the other side! Breakdown your complex model into small pieces has better and higher chances to enhance and achieve a high-resolution history match end-of-the-day especially with existing powerful Assisting History Match "AHM" tools & HPC's.
Senior Reservoir Engineer at Shell (PDO)
4 å¹´Thanks for sharing. What is your opinion on use of proxy equations on the AHM process? Rather than running numerous full physics simulation runs, "train" a proxy for the deemed observables and use this proxy equation to generate multiple realizations (parameters combinations) which feed your distribution curve. We normally have these discussions within our modelling forum but it is a difficult one to conclude. Run time of numerous full physics simulation vs. time spent to "train" the proxy. I would be interested to hear your thoughts.
Reservoir engineer | Digital and Integration | Simulation Modelling | Reservoir consultancy
4 年Thank you for sharing such valuable information, in fact that was the same approach that we followed more or less. The interesting thing, that we used different approach to build saturation hight model for each cell instead of building it from the available logs and populating them to rest of the structure using Arche equation. The sw that we estimated is function (porosity, permeability, HAFWL) so whenever you have porosity and permeability you can have a Pc curve and sw. And for the modifications that we made essentially were based on conventional reservoir engineering and according to that we created a map for modifying perm, baffles, barriers,fault transmissibility, Rel Perm ( end points for water break through and exponents for the trend of it). And this takes into account RFT data, CCA SCAL, well testing, pressure behavior along with the aquifer analysis using MBAL. It’s quite interesting!!