**IPLAB Solutions **

IPLab LLC offers the IP_Seismic software package for predicting the productivity parameters of oil and gas formations based on the use of machine learning algorithms and neural networks. The IP_Seismic software package has the ability for direct import/export data from/to Petrel software (Schlumberger). With other vendors like Kingdom (S&P Global), Landmark (Halliburton), Paradigm (AspenTech), etc data exchange is organaized via files exchange: seismic data through SEG-Y fies, wells through LAS-formats and horizons via text files. It is also pretty easy, user freindly and comfortable.

The Priezzhev Laboratory Limited Liability Company (IPLab LLC) is the sole copyright holder of the **IP_Seismic** software package, and has exclusive rights to grant the rights to use it, provide technical support services, training on it and other related services, as well as distribution on the territory of all countries of the world.

The exclusive right to the **IP_Seismic** software package was registered on April 20, 2017 with the federal executive authority for Intellectual Property of the Russian Federation in accordance with the procedure established by the Civil Code of the Russian Federation and other regulatory legal acts of the Russian Federation under the number 2017614592.

The **IP_Seismic** software package is registered in the Unified Register of Russian Computer Programs and Databases – the entry number in the register is 15221, the date of inclusion in the register is 14.10.2022, based on the order of the Ministry of Digital Development, Communications and Mass Communications of the Russian Federation dated 14.10.2022 according to the minutes of the Expert Council meeting dated 10.10.2022 No. 1483pr.

**Hardware recommendations**

Minimum recommended hardware for IP_Seismic package:

- Processor: Quad-core processor (best with a fast clock speed and high cache). Multi-core CPU recommended for parallel calculations.
- Memory: minimum 16 GB RAM / recommended 32 Gb RAM
- Graphics: any standard, GPUs are used to train Kolmogorov neural networks.
- Disk: recommended SSD to fast access to big seismic volumes
- Display: The quality of the viewing experience increases with the size and number of monitors

**Software requirements**

- Microsoft© Windows© Windows 10 or above 64-bit or Windows Server 64-bit
- Microsoft.NET Framework version 4.7 or above

**Installing **

To install the **IP_Seismic** package it is necessary just to unpack the installation **Zip** file to any folder.

Inside the folder there will be one **EXE** file – **IP_Seis.exe**, several **DLL** files and **Help** folder with documentations **PDF** files.

To start the package it is necessary to start **IP_Seis.exe**.

The license can be local or network.

At the first start, the program gives the **HOSTID** (local information about the CPU, disk, etc.) in error message:

The **HOSTID** (24 hexadecimal numbers) requires to be send to the email address: Priezzhev.I@ivanplab.ru

The license file will be generated and send back within one day. The file named **license.dat** must be placed to the same folder where **IP_Seis.exe**.

**Uninstalling**

You need just to delete the folder.

**Get started and Common description**

**Software package IP_Seismic:**

**I/O Modules**

**SEGY loader **– seismic data input in SEGY format,

**LAS loader **– input of borehole data in LAS,

**Excel wells loader **- input of well data in Excel format,

**Excel points attribute loader **- input of spatial point data with attributes in Excel format,

**ASCII points loader **- input of spatial point data with attributes in ASCII format,

**Surface ASCII loader **- input of 2D grid data with attributes in ASCII format,

**XYZ loader **- input of spatial mesh data with attributes in XYZ format,

**Well Marker loader **- input of markers data on wells in Petrel ASCII format,

**Time Depth log loader **- input of data "time-depth" on wells in ASCII format,

**SEGY export **– output of seismic data in SEGY format,

**Surface export **– output of 2D grid data with attributes,

**Points export **– output data points with attributes.

**Modules for working with 2D attributes (with a set of maps) **

**Maсhine Learning surface property prediction **– module for predicting effective parameters by a set of 2D attributes (a set of maps) based on linear regression, regression-based on neural networks, ACE regression, Random Forest regression, Nearest Neighbor regression, regression-based on neural networks with full function activation functions Kolmogorov type.

** At the input **– there are several maps with seismic attributes (for example, slicing amplitudes in the layer) and a set of intersection points with the predictive parameter (effective thickness, average porosity, cumulative production, ...)

** The result **– several maps with average forecasts, P10, P50, P90 and the forecast spread (standard deviation).

**Surface_Factor_Analysis **– a module for calculating independent factors by a set of 2D attributes (a set of maps) based on 1) Autoencoder technology - a "narrow neck"; and 2) PCA (Principal Component Analysis). It can be used to identify fracture zones.

** At the input **– there are several maps with seismic attributes (slicing amplitudes in the layer)

** The result **– several cards with independent components (factors).

**SOM classification **– a module for classifying a set of maps with a sliding window based on the self-organized mapping by Kohonen algorithm using 1D, 2D and 3D configurations of Kononen neural networks.

** At the input **– there are several maps with seismic attributes (slicing amplitudes in the layer)

** The result** – one or more maps with classification results.

**Calculator for surface properties **– a multi-line calculator based on c # with access to data in a sliding window

** At the input **– there are several maps with seismic attributes (slicing amplitudes in the layer)

** The result **– one card with the results of calculations.

**Modules for working with seismic cubes **

*As a result, the modules working with seismic cubes allow you to get virtual cubes. Such cubes retain only the computation algorithm and appear instantly. When visualizing on a cut, only those traces that are close to the cut are computed and therefore, it results in a possibility to visualize the result very quickly, without computing the entire cube. It also makes possible to correct “on-the-fly” the calculation parameters with fast visualization, which allows for a quick selection of parameters on a section with visualization of borehole curves. *

**Maсhine learning cube prediction **– a module for predicting effective parameters by a set of seismic cubes (full, angular amounts) based on linear regression, regression-based on neural networks, regression-based on Kolmogorov neural networks with lookup table activation functions.

** At the input **- one or several cubes (angular and azimuthal sums, any cubes with attributes) and several wells with predicted curves. It is possible to specify the roof and the bottom of the layer to limit the forecast.

** The result **– virtual cubes with forecast parameters, average forecasts, P10, P50, P90 and forecast spread (standard deviation).

**Poststack seismic inversion **– a module of inversion constructions for estimating an acoustic impedance based on classical inversion algorithms using a seismic pulse or on the extraction of a statistical pulse separately for each trace. The module also includes

** At the input **– there is one seismic cube (full amount). The pulse is automatically extracted from each trace.

** The result **– a virtual cube with AI, resulting in a quick change of parameters.

**AVO_seismic inversion **– a module of inversion constructions based on the joint application of a set of corner cubes before summation when calculating AI, Vp / Vs, density on the basis of Aki and Richards equations with calculation in the frequency domain.

** At the input **- a set of corner cubes, the pulse is extracted automatically from each trace. It is possible to specify a low-frequency model.

** The result **– virtual cubes with AI, Vp / Vs, density with the ability to change parameters quickly.

**AVOAZ_seismic inversion **– a module of inversion based on the joint use of a set of angular and azimuth cubes before summation aiming at calculation of AI, Vp / Vs, density, and the azimuthal heterogeneity indices? with calculation in the spectral domain.** Results **– virtual cubes with AI, Vp / Vs, density, and the indices of azimuthal density heterogeneity with the possibility of operative parameter change.** (Only for testing - not ready for commercial use) **

**Seismic_Factor_Analysis **– module for calculating independent factors for a set of 3D attributes (set of cubes) based on technology 1) Autoencoder - "narrow neck"; and 2) the method PCA (Principal Component Analysis). It can be used to identify fracture zones.** The result **– is several virtual cubes with orthogonal components.

**Cubes classification **– a 3D classification module for a set of seismic cubes based on the self-organized map (SOM) algorithm using 1D, 2D and 3D configurations of Kononen neural networks and RGB visualization of results. **The result **- is virtual cubes with classification results

**Seismic facies analysis **– a module for seismic facets in the form of a seismic signal in the layer under study, which uses a volumetric signal (sub-cube for several traces) for classification. As an algorithm for classification, the SOM (self-organized map) algorithm using 1D, 2D and 3D configurations of Kononen neural networks and RGB visualization of results is used.

**The result **– a set of maps with classification results.

**Calculator for cubes **– a multi-line calculator based on c # with access to data in a sliding window (an example of a procedure for computing a cube of coherence and calculating cubes for spectral decomposition is included). **The result **– virtual cube with the results of the calculation.

**Calculator for well logs **– a multi-line calculator based on c # for calculating new curves on a well with access to data in a sliding window.** The result **– a new curve with the results of the calculation.

**Modules of work with point data **

**Maсhine Learning cube prediction **– a module for predicting effective parameters for a set of attributes based on linear regression, regression based on neural networks, Random Forest regression, regression based on neural networks with arbitrary activation functions.** The result **– a new attribute with predictive parameters.

**Modules for working with gravimagnetic data **

**Layers gravy-mag modeling and inversion **– module for modeling and inversion of gravimagnetic data on the basis of 3D layered model. The inversion allows you to adjust the density or position of the boundaries of layers, taking into account the constraint model.

**Express gravy-mag modeling and inversion - **a module for modeling and inversion of gravimagnetic data based on a 3D model in the form of a cube SEGY.

**Visualization**

*Visualization modules include windows with visualization of maps, sections along lines or cross lines, arbitrary lines (polygones), crossplots, well section with logs and tops.*

*Fig. 1. Example of the IP_Seismic software interface.*

Figure 2. Above is a section along the seismic cube of the Groningen gas field. The section shows the projections of wells and the position of reflecting horizons. In the middle there is a section according to the predicted density cube. Below is a comparison of the measured and predicted density for several wells in the interval of the productive formation. A cross-raft of measured and predicted densities for all 366 wells. The correlation coefficient of the forecast is 0.8986.

Fig. 3. An example of a porosity cube prediction based on seismic field data and measurements in wells for seismic field drilling control. At the top is a section along the original seismic field. The graph on the horizontal well is the porosity curve. In the middle there is an incision along the porosity cube. The forecast was made using Kolmogorov neural networks. At the bottom there is a section along the porosity cube. The prediction was made using Kohonen neural networks using the "nearest neighbor" method.

There is a visualization of RGB maps, and RGB sections (mixing the colors of red,

green and blue for three cubes (real or virtual) or three cards.

Figure 4. An example of an RGB map based on three maps for several stratigraphic slices with an offset of 4 milliseconds between them. If the amplitudes change proportionally between the slices, then there will be a black-and-white picture. If color shades appear, it signals local changes in amplitudes.

Figure 4. An example of comparing the spectral decomposition map (on the left) and the RGB section on three slides with an offset of 2 milliseconds and with contrast based on the "Faults Simulation" algorithm.

Figure 5. An example of comparing an RGB map from a Western company (on the left) and an RGB map on three slides with an offset of 2 milliseconds and with contrast based on the "Faults Simulation" algorithm, calculated in IP_Seismic SW (on the right).