Search

CN-121980046-A - Snesim modeling-oriented search tree hybrid query acceleration optimization method and device

CN121980046ACN 121980046 ACN121980046 ACN 121980046ACN-121980046-A

Abstract

The invention discloses a Snesim modeling-oriented search tree hybrid query acceleration optimization method and device, the method comprises the steps of obtaining training images and data templates, constructing a search tree based on a Snesim method, constructing a reverse auxiliary search structure to establish a mapping relation between data event attribute values and search tree nodes, taking a simulation progress as a strategy switching index, adopting a reverse query strategy at a simulation initial stage of which the progress is smaller than a preset threshold value, reducing search tree traversal redundancy, and switching back to an efficient traditional forward query strategy at a middle-later stage of a simulation of which the progress is greater than or equal to the preset threshold value. The method can dynamically adapt to the evolution process of condition data from sparse to dense, obviously shortens the search query time, greatly improves the calculation efficiency and speed of Snesim modeling on the premise of guaranteeing the reproducibility of the geologic structure characteristics, and provides an effective optimization scheme for large-scale geologic modeling.

Inventors

  • YU SIYU
  • WEI KAILONG
  • LI SHAOHUA

Assignees

  • 长江大学

Dates

Publication Date
20260505
Application Date
20260114

Claims (10)

  1. 1. A Snesim modeling-oriented search tree hybrid query acceleration optimization method is characterized by comprising the following steps: Acquiring a training image and a data template, scanning the training image by using the data template, and constructing a search tree containing all data events and the repetition number thereof in the training image based on a Snesim method; Constructing a reverse auxiliary retrieval structure based on the search tree, and establishing a mapping relation among attribute values, hierarchical depths and node indexes of each node in the search tree; obtaining geometrical parameters and data files of a simulation grid, initializing the simulation grid based on the geometrical parameters, distributing data to corresponding nodes of the simulation grid, and defining random paths for accessing the nodes of the simulation grid; Judging whether the current simulation progress is smaller than a preset progress threshold according to the number of traversed and assigned nodes on the random path; If the simulation progress is smaller than the progress threshold, adopting a reverse query strategy, and searching the search tree nodes meeting the conditions by using the reverse auxiliary search structure, and counting the local conditional probability distribution of the nodes to be simulated; if the simulation progress is greater than or equal to the progress threshold, adopting a forward query strategy to start layer-by-layer matching data events from the root node of the search tree, and counting the local conditional probability distribution of the nodes to be simulated; And carrying out random sampling according to the local conditional probability distribution, and assigning the sampling result to the nodes to be simulated and taking the sampling result as the condition data of the subsequent simulation until all grid nodes are simulated.
  2. 2. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 1, wherein constructing a search tree containing all data events and the repetition number thereof in the training image based on Snesim method comprises: Initializing a search tree containing only root nodes; Sliding on the training image by utilizing the data template, and extracting all the data events; Inserting each extracted data event into the search tree one by one, and if the path corresponding to the data event exists, increasing the repetition count of the corresponding node; If no path exists, a new node is created and a repetition count is initialized until all data events within the training image based on the data template are inserted into the search tree.
  3. 3. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 1, wherein constructing a reverse auxiliary search structure based on the search tree comprises: Creating a dictionary set corresponding to each level of the search tree; traversing each node of the search tree, and acquiring a corresponding attribute class value of the node in the training image and the hierarchical depth of the node in the search tree; And taking the attribute category value as a key, taking the set of all the search tree nodes which have the same attribute category value and are positioned at the same hierarchical depth as a value, and storing the set of all the search tree nodes into a dictionary of the corresponding hierarchy.
  4. 4. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 2, wherein constructing a search tree containing all data events and the repetition number thereof in the training image based on Snesim method further comprises: Assigning a node path code to each node in the search tree, the node path code being used to identify a path from a root node to the node to represent a sequence of data events corresponding to the node; the reverse-aiding retrieval structure is also used to store the node path codes for each node.
  5. 5. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 1, wherein the reverse query strategy comprises: selecting a condition point farthest from a node to be simulated from the data event; Based on the level and the value of the condition points in the search tree, reading the node path code of each candidate node in the set through a reverse auxiliary search structure, and judging whether the attribute value sequence represented by the node path code is consistent with the attribute values of other known condition data points except the farthest condition point in the current data event; Candidate nodes with consistent attribute values are screened out, the repetition number of the attribute values of the central position of the nodes is counted, and the local conditional probability distribution is calculated.
  6. 6. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 5, wherein the reverse query strategy further comprises: And deleting the current furthest condition point and re-executing reverse query until the matched node is obtained or all the condition points are deleted when no candidate tree nodes meeting all the condition point constraints exist.
  7. 7. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 1, wherein the forward query strategy comprises: Initializing a current search tree node as a root node; determining whether the current data event has a known condition data point at a location corresponding to a next level of a search tree; If the known condition data point exists, only retaining the child node with the same attribute value as the known condition data point as the current search tree node of the next level; if the known condition data points do not exist, all child nodes of the current node are reserved as candidate paths of the next level; and executing the judging process until the leaf nodes of the search tree are reached or the matching cannot be continued, and counting the repetition number of the central attribute values of all the end nodes of the reserved paths.
  8. 8. The Snesim modeling oriented search tree hybrid query acceleration optimization method of claim 7, wherein the forward query strategy further comprises: And when the number of the matching paths obtained by forward query is smaller than a preset threshold value, performing template simplification operation on the data event so as to increase the number of the matching paths.
  9. 9. A Snesim modeling-oriented search tree hybrid query acceleration optimization device, comprising: the search tree construction module is used for acquiring a training image and a data template, scanning the training image by utilizing the data template, and constructing a search tree containing all data events and the repetition number thereof in the training image based on a Snesim method; the reverse auxiliary retrieval structure construction module is used for constructing a reverse auxiliary retrieval structure based on the search tree and establishing a mapping relation among attribute values, hierarchical depth and node indexes of each node in the search tree; The device comprises a setting module, a setting module and a control module, wherein the setting module is used for acquiring geometric parameters and data files of a simulation grid, initializing the simulation grid based on the geometric parameters, distributing data to corresponding nodes of the simulation grid, and defining a random path for accessing the nodes of the simulation grid; The strategy control module is used for judging whether the current simulation progress is smaller than a preset progress threshold according to the number of the traversed and assigned nodes on the random path, adopting a reverse query strategy if the simulation progress is smaller than the progress threshold, searching search tree nodes meeting the conditions by using the reverse auxiliary search structure, and counting the local conditional probability distribution of the nodes to be simulated; and the probability statistics and assignment module is used for randomly sampling according to the local conditional probability distribution, assigning the sampling result to the nodes to be simulated and taking the sampling result as the condition data of the subsequent simulation until all the grid nodes are simulated.
  10. 10. A computer readable storage medium storing a computer readable program or instructions which, when executed by a processor, is capable of carrying out the steps of the Snesim-modeling oriented search tree hybrid query acceleration optimization method of any one of claims 1-8.

Description

Snesim modeling-oriented search tree hybrid query acceleration optimization method and device Technical Field The invention relates to the technical field of geostatistical modeling, in particular to a Snesim modeling-oriented search tree hybrid query acceleration optimization method and device. Background Multipoint geostatistics (Multiple-point Geostatistics, MPS) can effectively characterize the spatial heterogeneity of a reservoir, quantitatively characterize the distribution characteristics and stacking relationships of various lithofacies in three-dimensional space. With the continued depth of oil and gas exploration and development, the number of simulated grid cells required for geologic modeling has reached tens of millions of scales in order to more finely reveal reservoir internal structures, which also means that the simulation process presents significant challenges in terms of computation time and memory usage. Relevant expert scholars are working on optimization and improvement of multipoint geostatistical algorithms to improve modeling efficiency. Currently, MPS methods can be generally classified into pixel-level algorithms and block-level algorithms according to the number of mesh nodes per simulation. The pixel level algorithm is represented by Snesim, and the efficiency of the early pixel level algorithm is remarkably improved by constructing a conditional probability search tree. However, when processing training images with long-distance spatial correlation features, if the template size is too large, the size of the search tree increases rapidly, resulting in a large increase in memory usage and computation, and limited efficiency. Block level algorithms are represented by SIMPAT and DIRECT SAMPLING (DS). Block level algorithms are generally more efficient than pixel level algorithms by modeling a batch of nodes at a time. However, SIMPAT has the core problems that the simulation efficiency shows serious unbalance and unbalance in the whole process, the algorithm has acceptable speed in the initial stage of simulation (sparse condition data), but along with the promotion of simulation, condition data around the nodes to be simulated become highly dense, the traditional query mechanism has to perform deep and complex layer-by-layer matching and screening in a huge search tree so as to meet the constraint of all dense conditions, the calculation amount and time consumption of unit nodes are increased in a nonlinear way, the speed in the later stage of simulation is greatly reduced, the efficiency of large-scale and high-precision geologic modeling is severely restricted, and the DS algorithm has the advantage of dynamically adjusting the neighborhood scale, but the single-step simulation needs to scan most part of training images, and the calculation efficiency still limits the large-scale application. In addition, in practical application, factors such as selection of a simulation path and a retrieval mechanism of a training image mode also have significant influence on the calculation efficiency of the existing algorithm. Along with the development of fine geologic modeling technology and the improvement of large-scale and high-resolution modeling requirements, snesim gradually shows a certain restriction in terms of calculation efficiency, and becomes an important factor for restricting further wide application. Therefore, a new comprehensive conditional probability query method is required to be provided, and the calculation efficiency is improved on the basis of keeping the excellent statistical feature retention capability of Snesim algorithm and the collaborative prediction function of soft and hard data, so as to meet the requirement of fine geologic modeling on large-scale and high-resolution simulation. Disclosure of Invention In view of the above, the invention provides a Snesim modeling-oriented search tree hybrid query acceleration optimization method and device, which are used for solving the technical problems that the existing Snesim modeling method is low in query efficiency and slow in overall modeling efficiency when initial condition data are sparse in simulation, and cannot meet the current large-scale geological modeling requirements. In order to achieve the technical purpose, the invention adopts the following technical scheme: In a first aspect, the present invention provides a method for accelerating optimization of a search tree hybrid query for Snesim modeling, including: Acquiring a training image and a data template, scanning the training image by using the data template, and constructing a search tree containing all data events and the repetition number thereof in the training image based on a Snesim method; Constructing a reverse auxiliary retrieval structure based on the search tree, and establishing a mapping relation among attribute values, hierarchical depths and node indexes of each node in the search tree; obtaining geometrical parameters and data fi