Multi-system compatible fully-sensing self-optimizing power grid project digital twin model and its data alignment method

Abstract

The present invention discloses a multi-system compatible fully-perceived self-optimizing power grid project digital twin model and its data alignment method. The transformation alignment of BIM model and GIS data is based on linear affine transformation to construct the underlying algorithm and perform multi-parameterized local nonlinear optimization matching to obtain the main alignment algorithm framework; at the same time, based on the self-developed data fusion technology, it is integrated with the power grid GIS (or other power grid data system) to simulate the full life cycle data of the power construction project to build the consistent coordination and comprehensive perception of the power construction data. The present invention has carried out secondary development based on the power construction BIM model and the power grid GIS system, and proposed an efficient BIM model and GIS data geometric transformation and alignment method, which achieves data alignment with higher accuracy and efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of digitalization of power grids, and in particular to a multi-system compatible fully-sensing self-optimizing power grid project digital twin model and a data alignment method thereof. Background Art

[0002] In the field of power grid construction projects and the operation and maintenance of power construction projects, digital transformation is becoming a key driving force for improving efficiency, ensuring safety and optimizing resource allocation. Existing technologies, such as building information modeling (BIM), geographic information systems (grid GIS), real-time monitoring and data acquisition systems, have been widely used in the planning, design, construction and operation and maintenance stages of power grid projects. These technologies provide the capabilities of three-dimensional visualization, spatial data analysis and real-time monitoring, but they often exist as isolated systems and lack effective data integration and collaborative working mechanisms.

[0003] Under the current technological background, the construction and optimization of the digital twin model of the power grid is one of the key links in the digital transformation of the power system. This process involves highly complex data integration and processing, especially the effective integration of building information model (BIM) and geographic information system (GIS) data. Existing technologies mainly rely on linear transformation and simple data matching methods to achieve the alignment of BIM models and GIS data, but these methods often cannot fully cope with the geometric and attribute differences between the two, resulting in unsatisfactory accuracy and efficiency of data alignment; especially when faced with more complex nonlinear data deformations, their performance and accuracy are greatly reduced. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a technical framework for a digital twin model system of a fully-perceived self-optimizing power grid project, and at the same time to deeply develop a data alignment algorithm compatible with multiple systems therein.

[0005] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0006] The geometric transformation and alignment method of BIM model and GIS data in the digital twin model of power grid is used for the secondary development of BIM model in the construction of digital twin model of power grid. The transformation alignment of BIM model and GIS data is based on linear affine transformation to construct the underlying algorithm, and multi-parameterized local nonlinear optimization matching is performed to obtain the main alignment algorithm framework; and the optimal dynamic transformation alignment of BIM model and GIS data is achieved through subsequent iterative optimization based on the model of the main alignment algorithm framework as needed.

[0007] As a preferred technical solution of the present invention, the main alignment algorithm framework and its iteratively optimized data process route are as follows: first, feature points are extracted and matched; feature points are extracted from the BIM model and GIS data; the feature points are constructed to point to corners, edges or other significant geometric or geographic markers; corresponding feature points between the BIM model and the GIS data are found, and feature descriptors based on mathematical descriptions of the areas around the feature points are used to match the feature points in the BIM model and the GIS data, wherein the feature descriptors are constrained based on rotation and scale invariance; the similarity of the feature descriptors is compared to find the best matching feature point pair and optimize it; and a preliminary affine transformation is further performed. Estimation; The basic algorithm structure of linear affine transformation is set to include two sets of parameters: coefficient parameters and translation parameters; Based on this basic data structure, the first step is to introduce the least squares method to construct a linear equation group for estimating the affine transformation parameters and obtain the initial binary parameters by eliminating the equation group; Then, local nonlinear transformation parameter estimation is performed; With the above-mentioned affine transformation results as input, the geometric data deformation analysis tool is introduced into the main alignment algorithm framework to construct a measurement function including data deformation energy and data terms, and the corresponding linear data system is derived and solved to find the transformation parameters to minimize the measurement function, and the local nonlinear transformation parameters are obtained and subsequently optimized as needed.

[0008] As a preferred technical solution of the present invention, the main alignment algorithm framework and the specific data process of its iterative optimization include:

[0009] Step 1: Feature point extraction and matching; Existing computer vision algorithms are used to extract feature points from BIM models and GIS data. The following algorithmic data process is constructed for the matching of feature points: In order to achieve accurate matching of feature points between BIM models and GIS data, key points and their feature descriptions are first extracted based on the scale-invariant feature transformation algorithm. Key points are detected at different pre-specified scales or global scales and a feature descriptor is generated for each key point. The descriptor is consistent and invariant to rotation and scaling, so that it can be used to match between different data and their geometric representations; Next, the feature descriptors extracted from the BIM model are matched with the feature descriptors in the GIS data. The data distance is used as a measure of the difference between the two feature descriptors. The data square sum or square root distance between the feature descriptors in each BIM model and all feature descriptors in the GIS data is calculated, and it is matched with the nearest corresponding data. Figuratively speaking, this data process is to find the most matching data partner in the GIS data for each key point in the BIM model;

[0010] As a preferred technical solution of the present invention, the next step is to perform subsequent directional matching optimization to improve the accuracy and robustness of the matching. Specifically, two optimization strategies are constructed: α, a ratio test to exclude unreliable matches. On the basis of the distance between the feature descriptor in each BIM model and the nearest neighbor in the GIS data, the distance to the next nearest neighbor is added. If the distance ratio between the nearest neighbor and the next nearest neighbor is less than a preset threshold, that is, a value less than 1, the match is considered to be valid, thereby excluding ambiguous matches and improving the reliability of the matching; β, a geometric consistency check. This check process is constructed to improve the consistency of the geometric relationship of the data between the matching points. The data process first estimates the initial transformation model, that is, the affine transformation adopted, and verifies the proportion of all matching points that conform to the transformation model by randomly selecting a small number of matching points. On this basis, the largest set of matching points that conform to it is obtained through repeated iterations.

[0011] As a preferred technical solution of the present invention, step 2: preliminary affine transformation estimation; the basic algorithm structure of the linear affine transformation is set to include two groups of parameters: coefficient parameters and translation parameters; specifically, the basic algorithm structure of the linear affine transformation is set to: X' = AX + b; wherein X is the original coordinate, A is a 3×3 transformation matrix (the data space in BIM and GIS is three-dimensional; the spatial dimension can be adjusted as needed), b is the translation vector, and X' is the transformed coordinate; on this basic data structure, the first step is to introduce the least squares method to construct a linear equation group for estimating affine transformation parameters A and b, and obtain the initial binary parameters by eliminating the equation group; further, construct an optimization process for preliminary affine transformation estimation, and since the preliminary affine transformation estimation is a linear transformation, the existing RANSAC algorithm is directly used here to eliminate erroneous matches to optimize the transformation parameters;

[0012] As a preferred technical solution of the present invention, step 3: local nonlinear transformation parameter estimation; specifically, the geometric data deformation analysis tool is introduced into the main alignment algorithm framework to optimize the local nonlinear transformation parameters. Both BIM and GIS data naturally have geometric space attributes, and the local nonlinearity of the data can be vividly imported and corresponded with the geometric shape transformation and data alignment; the data processing of the geometric shape transformation and data alignment finds the best deformation mapping by minimizing a measurement function; at this time, firstly, a measurement function E(f) including data deformation energy and data terms is constructed according to the following data standards. This function consists of two parts: one part is the deformation energy, which is used to punish excessive deformation; the other part is the data term, which ensures that the transformed point is as close to the target point as possible; based on this measurement function, the smoothness of the data deformation of the two system components of BIM and GIS and the correspondence between the data points are balanced; on this basis, a method for characterizing data is constructed according to the following standards. The transformation function of the transformation process is set to include two sub-items: the linear transformation sub-item corresponds to the data representation of the rotation and / or scaling and / or translation operation, and the nonlinear deformation sub-item corresponds to the data representation of the local geometric adjustment, thereby finely adjusting the local details while maintaining the overall structure; at this time, in order to find the optimal transformation parameters based on the linear affine transformation, the problem of minimizing the measurement function is converted into a linear system problem. Specifically, the data processing process is set to set the partial derivative of the measurement function with respect to each parameter to zero, thereby obtaining a set of linear equations, and this linear equation group is associated with the parameters including the linear transformation sub-item and the nonlinear deformation sub-item; in this way, after obtaining the linear system represented by the linear equation group, the numerical method is directly embedded for solution, such as embedding LU decomposition or conjugate gradient method to solve the offline system to obtain a set of optimal parameters so that the measurement function reaches the minimum value and the corresponding optimal deformation mapping; and further construct the subsequent optimization data process as needed.

[0013] As a preferred technical solution of the present invention, in step 3, the subsequent data optimization process includes: the optimization strategy of local nonlinear transformation is constructed as an application feedback model, specifically, the above-mentioned local nonlinear transformation function is applied to any point in the BIM and / or GIS source data space using the solved parameters, and its position in the GIS and / or BIM target data space is calculated, and the linear and nonlinear alignment (especially nonlinearity) of the source data to the target data, the proximity of the transformed point to the target point, and the smoothness of the data deformation are monitored and compared; the feedback data model is iterated and optimized according to the monitoring and comparison results.

[0014] The reference system transformation and alignment method of BIM model and GIS data in the power grid digital twin model regards the different source data in the power construction BIM model and the power grid GIS system as different reference systems, and transforms the data alignment into finding the invariant in different reference systems, that is, the corresponding relationship between the two data sets in the space where the BIM model and the GIS system are located, and achieves the best data alignment by adjusting the spatiotemporal structure of the two data spaces, that is, the spatial position of the data points.

[0015] A continuous manifold alignment method for BIM model and GIS data in the power grid digital twin model is proposed. Data modeling is performed based on the differential geometry continuous manifold after data interpolation to align data and realize the fusion of BIM model and GIS data.

[0016] A fully-sensing self-optimizing digital twin model system for power grid projects. This model system is based on BIM model technology for further multi-system data fusion secondary development. When establishing the BIM model, geometric modeling and computer graphics technology are used to create an accurate geometric three-dimensional model. At the same time, based on the self-developed data fusion technology, it is integrated with the power grid GIS (or other power grid data system), and the consistent coordination and comprehensive perception of power construction data are constructed for the full life cycle data simulation of the power construction project.

[0017] As a preferred technical solution of the present invention, the key steps of constructing the model system include:

[0018] Data specification and standardization: formulate unified data specifications and data standards for power construction projects to ensure data consistency and interoperability; determine data classification, coding, format and interface standards to facilitate data integration and sharing;

[0019] BIM model integration construction and digital twin platform development: Use BIM technology to create a three-dimensional model of the power construction project, including buildings, equipment, and pipelines; integrate the BIM model with the power grid GIS geographic information system and / or the power grid monitoring and data acquisition system and / or other necessary or required systems to achieve the integration of spatial data and real-time monitoring data; Among them, the integration of the BIM model and the power grid GIS geographic information system is based on the geometric transformation and alignment method of the BIM model and GIS data; Then, based on the above integrated BIM model, a digital twin platform that supports multi-dimensional data integration and real-time data processing is constructed, and supports real-time update and dynamic visualization of data;

[0020] Full-chain data integration: Integrate data from all aspects of power construction project planning, design, construction, and operation and maintenance, including but not limited to drawings, plans, progress, costs, quality, and safety information, and realize real-time flow and update of data;

[0021] Development of self-optimizing algorithms: Based on data analysis and machine learning, self-optimizing algorithms are built to predict and optimize the performance and cost of power construction projects. The algorithms automatically adjust the operating parameters of power construction projects based on historical and real-time data.

[0022] Establishment of a digital collaborative ecosystem: Establish an internal and external digital collaborative ecosystem based on power construction project participants, suppliers, and regulatory agencies, and achieve data sharing and collaboration through application programming interfaces (APIs) and data sharing protocols;

[0023] Data sharing and collaborative support: Build a set of data sharing mechanisms to coordinate the access and use of required data by power construction project participants and support collaborative interactions in the construction of power grid power projects, including design collaboration, construction collaboration, and operation and maintenance collaboration.

[0024] As a preferred technical solution of the present invention, for the data specification and standardization steps, statistical tools are introduced to analyze the characteristics of data and establish a set of unified data standards, including data collection, processing, storage and exchange standards;

[0025] As a preferred technical solution of the present invention, for the development step of the self-optimization algorithm, the development of the self-optimization algorithm is based on the technology of machine learning and artificial intelligence, using supervised learning and unsupervised learning methods to predict and optimize the performance of power construction projects; considering building a dedicated neural network to establish a complex local nonlinear model and using a genetic algorithm to find the optimal solution;

[0026] As a preferred technical solution of the present invention, for the steps of establishing a digital collaborative ecosystem, a system analysis is conducted based on the architectural design of the information system and network theory is used to optimize the process of data sharing and collaborative work; for data sharing and collaborative support, encryption algorithms, data compression technologies or network communication protocols are introduced or constructed with the goal of allowing data to be securely shared between different systems and organizations.

[0027] As a preferred technical solution of the present invention, the model system also includes the following optional subsequent construction steps: secondary customization of multiple business scenarios, customizing the application scenarios of digital twin models including asset management, risk assessment, and fault diagnosis according to different business needs; and providing a flexible user interface and reporting tools through data multi-channel ports;

[0028] As a preferred technical solution of the present invention, the model system also includes the following subsequent optional construction steps: security and privacy protection, taking data security and privacy protection into consideration in system design, and taking encryption and access control measures; at the same time, performing data legalization self-screening with manual supervision or manual intervention based on a pre-imported legal and regulatory database.

[0029] The beneficial effect of adopting the above technical solution is that the present invention has carried out secondary development based on the electric power construction BIM model and the power grid GIS system, and proposed an efficient BIM model and GIS data geometric transformation and alignment method, which is based on the underlying algorithm of linear affine transformation and combined with multi-parameterized local nonlinear optimization matching to achieve higher precision and efficiency data alignment. The details of the technical effect of the present invention can be found in the specific embodiments below. DETAILED DESCRIPTION

[0030] Example 1: Overview of Digital Twin Model Development

[0031] The construction of the digital twin model of the power construction project is a systematic interdisciplinary research project. After the project was established within the power grid, a technical team was assembled, including internal power engineers, architects, software engineers and data experts from Langxin Company, and scientific researchers from universities. At the same time, the power construction project management team, construction team, operation and maintenance team, etc. were also formed to work closely together.

[0032] In terms of development route guidance, the specific construction of the digital twin model of the power construction project is based on the consistent data specifications and data standards of the power construction project, integrating data information from the entire chain and all links to achieve the unification of data sources for the power construction project, and comprehensively perceive the full life cycle status of power grid infrastructure and retail power construction projects; at the same time, collaboratively establish an internal and external digital collaborative ecosystem for the power construction project to achieve data sharing and multi-data collaboration for the power construction project.

[0033] In terms of technical route guidance, the model system is based on BIM model technology and conducts further secondary development of multi-system data fusion. When establishing the BIM model, geometric modeling and computer graphics technology are used to create accurate geometric three-dimensional models. At the same time, it is integrated with the power grid GIS (or other power grid data systems) based on self-developed data fusion technology, and builds consistent coordination and comprehensive perception of power construction data for full life cycle data simulation of power construction projects.

[0034] Example 2: Model data integration (data alignment) technical route

[0035] BIM model fusion construction is a core development project of the digital twin platform, which requires integrating the basic BIM model with the power grid GIS or other power grid data systems based on self-developed data fusion technology.

[0036] Based on technical group discussions, a variety of data alignment solutions were initially proposed, including:

[0037] The ω scheme and the invariant model regard different data sources as different reference frames and use the reference frame transformation method to align the data;

[0038] The ψ scheme, based on the idea of ​​data geometry, integrates multi-dimensional data into the geometric manifold model for differential processing;

[0039] The χ scheme is also a composite data processing route based on linear affine transformation + local nonlinear parameter optimization under the data geometry technology route.

[0040] Finally, the χ scheme was selected as the final technical development route. Specifically, under the composite data processing route based on linear affine transformation + local nonlinear parameter optimization, the transformation alignment of BIM model and GIS data is based on the construction of the underlying algorithm based on linear affine transformation, and multi-parameterized local nonlinear optimization matching is performed to obtain the main alignment algorithm framework; and the optimal dynamic transformation alignment of BIM model and GIS data is achieved through subsequent iterative optimization based on the model of the main alignment algorithm framework as needed.

[0041] Example 3, technical implementation plan of route χ

[0042] The basic algorithm idea of ​​the χ scheme is as follows: for the main alignment algorithm framework, firstly, a basic affine transformation is constructed, and this basic affine transformation is constructed as a linear transformation; further, for the local nonlinear matching problem of the BIM model and the data in the GIS system, it is corresponded to the geometric local deformation by data model, and under this data guidance, multi-parameterized local nonlinear optimization matching is performed to obtain the main alignment algorithm framework; for the optimization of the main alignment algorithm framework, firstly, a specific optimization process and data target are constructed, and then iterative optimization is performed based on feedback supervision or machine learning of artificial standardized data. In terms of technical route, the main alignment algorithm framework needs to first extract feature points from the BIM model and GIS data, and then embed the corresponding feature points between the BIM model and GIS data, that is, match the feature points. On this basis, the coefficient parameters and translation parameters of the linear affine transformation are obtained based on the matching feature point fitting, thereby completing the preliminary affine transformation alignment; the transformation of the BIM model and GIS data feature points corresponding to the local nonlinearity after completing the preliminary affine transformation alignment is non-aligned, and the size of the deformation corresponds to the degree of local nonlinearity of the original data feature points. Therefore, the preliminary affine transformation result is further used as input to construct a minimization index representation composed of multi-parameter transitions, and the geometric deformation of the data is minimized by optimizing or solving the minimization index; the algorithmic technical route of iterative optimization of the main alignment algorithm framework is: based on data algorithm simulation and / or application evaluation of the transformation model, the feature point matching and transformation parameters are iteratively adjusted based on the quality of the BIM model and the data in the GIS system, so as to achieve the best dynamic transformation alignment of the BIM model and GIS data.

[0043] Specifically, the data process and its basic algorithm framework under the χ scheme are as follows:

[0044] Step 1: Feature point extraction and matching: Use computer vision algorithms to automatically identify feature points in the BIM model and GIS data. These algorithms work by detecting corners, edges, or other significant features in the image; such as building corners, window edges, intersections, terrain landmarks, etc. Then find the corresponding feature points between the BIM model and the GIS data. Here, we mainly seek to achieve high-accuracy feature point matching between the BIM model and the GIS data, and seek to supplement the construction of matching optimization strategies to improve the accuracy and robustness of the matching.

[0045] The first aspect of seeking: Specifically, in order to achieve accurate matching of feature points between BIM models and GIS data, key points and their feature descriptions are first extracted based on the scale-invariant feature transformation algorithm, key points are detected at different pre-specified scales or global scales, and a feature descriptor is generated for each key point. The descriptor is consistently invariant to rotation and scaling, so that it can be used to match between different data and their geometric representations; next, the feature descriptors extracted from the BIM model are matched with the feature descriptors in the GIS data, and the data distance is used as a measure of the difference between the two feature descriptors. The data square sum or its square root distance between the feature descriptors in each BIM model and all feature descriptors in the GIS data is calculated, and matched with the nearest corresponding data. Figuratively speaking, this data process is for each key point in the BIM model to find the most matching data partner in the GIS data. Finally, the basic construction of the algorithmic data structure framework for this first aspect is carried out as follows: Feature extraction: Use scale-invariant feature transform SIFT to detect key points and generate invariant feature descriptors at different scales. For each key point i, the algorithm extracts its feature descriptor d_i ∈ R^128, and the latter's data structure is a multidimensional vector, such as 128 dimensions; Feature matching: Use Euclidean distance as the distance measure between feature descriptors; For the feature descriptor d_i in the BIM model and the feature descriptor d_j in the GIS data, the distance between them is defined as: D(d_i, d_j) = sqrt(∑(k=1)^128 (d_i[k] - d_j[k])^2); Nearest neighbor matching: For each feature descriptor d_i in the BIM model, find the nearest neighbor d_j in the feature descriptor set of GIS data, that is: j = arg min_j D(d_i, d_j).

[0046] The second aspect is to conduct subsequent directional matching optimization to improve the accuracy and robustness of matching. Specifically, two optimization strategies are constructed: α, a ratio test to exclude unreliable matches. On the basis of the distance between the feature descriptor in each BIM model and the nearest neighbor in the GIS data, the distance to the next nearest neighbor is added. If the distance ratio between the nearest neighbor and the next nearest neighbor is less than a preset threshold, that is, a value less than 1, the match is considered valid, thereby excluding ambiguous matches and improving the reliability of matching; β, geometric consistency test. This test process is constructed to improve the consistency of the geometric relationship between the matching points. The data process first estimates the initial transformation model, that is, the affine transformation adopted, and verifies the proportion of all matching points that conform to the transformation model by randomly selecting a small number of matching points. On this basis, the largest set of matching points that conform to it is obtained through repeated iterations. At the same time, the algorithmic data structure framework for the second aspect of the search was also constructed as follows: The matching optimization strategy tentatively introduced ratio tests and consistency tests; Ratio test: For each matching pair (d_i, d_j), find the nearest neighbor d_j and the second nearest neighbor d_j' of d_i in the BIM model. If the ratio of the nearest neighbor distance to the second nearest neighbor distance is less than a certain threshold τ (for example, 0.8), the match is considered valid: D(d_i, d_j) / D(d_i, d_j') <τ. Geometric consistency check: According to the technical route, the random sampling consistency algorithm is introduced as a mathematical framework.

[0047] Step 2: Preliminary affine transformation estimation; The basic algorithm structure of the linear affine transformation is set to include two sets of parameters: coefficient parameters and translation parameters; Specifically, the basic algorithm structure of the linear affine transformation is set to: X' = AX + b; Where X is the original coordinate, A is a 3 × 3 transformation matrix, b is the translation vector, and X' is the transformed coordinate; Based on this basic data structure, the first step is to introduce the least squares method to construct a linear equation group for estimating the affine transformation parameters A and b, and obtain the initial binary parameters by eliminating the equation group; Further, in order to improve the robustness, the RANSAC algorithm is used to eliminate incorrect matches and find the best transformation parameters;

[0048] Step 3: Local nonlinear transformation parameter estimation: Taking the affine transformation result of the previous step as input, the next step is to introduce the geometric data deformation analysis tool into the main alignment algorithm framework, construct a measurement function E(f) that includes data deformation energy and data terms, and derive and solve the linear data system to find the transformation parameters that minimize the measurement function, obtain the local nonlinear transformation parameters and perform subsequent optimization as needed.

[0049] Here, attention should be paid to balancing the smoothness of deformation and the correspondence between data points to achieve accurate and smooth data alignment between data points of different systems. Specifically, the geometric data deformation analysis tool is introduced into the main alignment algorithm framework to optimize the local nonlinear transformation parameters. Both BIM and GIS data naturally have geometric space attributes. The local nonlinearity of the data can be vividly imported and corresponded with the geometric shape transformation and data alignment; the data processing of geometric shape transformation and data alignment finds the best deformation mapping by minimizing a measurement function; at this time, firstly, a measurement function E(f) containing data deformation energy and data terms is constructed according to the following data standards. This function consists of two parts: one part is the deformation energy, which is used to punish excessive deformation; the other part is the data term, which ensures that the transformed point is as close to the target point as possible; based on this measurement function, the smoothness of the data deformation and the correspondence between the data points of the two systems BIM and GIS are balanced; on this basis, a transformation used to characterize the data transformation process is constructed according to the following standards Function, the transformation function is set to include two sub-items: the linear transformation sub-item corresponds to the data representation of rotation and / or scaling and / or translation operations, and the nonlinear deformation sub-item corresponds to the data representation of local geometric adjustment, thereby finely adjusting the local details while maintaining the overall structure; at this time, in order to find the optimal transformation parameters based on the linear affine transformation, the problem of minimizing the measure function is transformed into a linear system problem. Specifically, the data processing process is set to set the partial derivative of the measure function with respect to each parameter to zero, thereby obtaining a set of linear equations, and this linear equation group is associated with the parameters including the linear transformation sub-item and the nonlinear deformation sub-item; in this way, after obtaining the linear system represented by the linear equation group, a numerical method is directly embedded for solution, such as embedding LU decomposition or conjugate gradient method to solve the offline system to obtain a set of optimal parameters so that the measure function reaches the minimum value and the corresponding optimal deformation mapping.

[0050] In the subsequent data optimization process, the optimization strategy of local nonlinear transformation is constructed as an application feedback model. Specifically, the above-mentioned local nonlinear transformation function is applied to any point in the BIM and / or GIS source data space using the solved parameters, and its position in the GIS and / or BIM target data space is calculated. The linear and nonlinear alignment (especially nonlinearity) of the source data to the target data, the proximity of the transformed point to the target point, and the smoothness of the data deformation are monitored and compared; the feedback data model is iterated and optimized according to the monitoring and comparison results.

[0051] The algorithmic data structure frame of this step includes the following basic parts: The measure function is represented as: E(f) = α ∫Ω |Δf|^2 dΩ + ∑(i=1)^n |f(p_i) - q_i|^2. Among them, α is the smoothing parameter, which controls the balance between deformation energy and data term; Δf is the Laplacian of f, which indicates the smoothness of the deformation; p_i is the point in the source space, q_i is the corresponding point in the target space; n is the number of point pairs. Here, the measure function E(f) consists of two parts: deformation energy and data term; the deformation energy penalizes excessive nonlinear deformation of the data, while the data term ensures that the transformed point is as close to the target point as possible. Transformation function TPS; the transformation function f is represented as a combination of an affine part and a nonlinear deformation part: f(x) = Ax + b + ∑_(j=1)^n w_j φ(|x - p_j|). Where A is a linear transformation matrix, b is the translation vector, w_j is the weight coefficient, and φ(r) = r^2 log r^2 is the radial basis function used to describe the nonlinear deformation between points x and p_j. A linear system is constructed for the transformation parameters (A, b, and w_j), and the optimization problem corresponding to the measurement function E(f) is transformed into a linear system; by setting the partial derivative of E(f) with respect to each parameter equal to zero, a linear system of equations is obtained: [K + α IP ] [ P^T 0][ w ] [ v ]= [ Q ] [ 0 ]; where K is the kernel matrix, whose element K_ij = φ(|p_i - p_j|); I is the identity matrix; P is the matrix consisting of the source point coordinates, with an additional row of all 1s to represent the translation; w is the weight vector; v contains the affine transformation parameters; and Q is the target point coordinate matrix. Then, LU decomposition or conjugate gradient method is introduced to solve the above linear system to obtain the transformation parameters (A, b, and w_j) and apply, feedback, and optimize.

[0052] The data flow in this basic algorithm frame is: sufficient feature point pairs as the basis for subsequent transformations → a preliminary global alignment (linear affine) lays the foundation for more refined local alignment → further optimize the alignment through local nonlinear transformations, allowing local adjustments to adapt to the differences between the two data models BIM and GIS; as well as subsequent transformation applications and data alignment implementations, and ensuring the quality of the alignment results through evaluation and adjustment. The data flow in the basic algorithm frame shows a step-by-step refined alignment strategy from global to local, each step is based on the results of the previous step, and provides the necessary input for the next step, ensuring the coherence and effectiveness of the entire process.

[0053] Example 4, Overview of ω Scheme and ψ Scheme

[0054] The ω scheme regards different source data in BIM and GIS as different reference systems, and uses the reference system transformation method to align data. Its data model is simpler and clearer (but the reliability of data processing quality may be slightly worse). The ω scheme regards the data alignment problem as finding an "invariant" (i.e. the correspondence between the two data sets) in different "reference systems" (i.e. the space where the BIM model and GIS data are located), and achieves the best alignment by adjusting the "space-time structure" (i.e. the spatial position of the data points).

[0055] The basic technical idea of ​​the ω scheme. First, define "spacetime" and "invariant". "Spacetime" definition: The space where the BIM model and GIS data are located is regarded as "spacetime", where the position of the data point is equivalent to the event in spacetime. "Invariant" definition: The correspondence between the BIM model and GIS data is defined as "invariant", that is, the correspondence between a specific BIM model point and a GIS data point should remain unchanged in any reference system. Then, the concept of "BG transformation" is introduced. In the special theory of relativity, the transformation describes the conversion of spacetime coordinates between different inertial reference systems. Drawing on this concept, a similar transformation can be designed to adjust the position of the BIM model point to achieve the best alignment with the GIS data; specifically, a Lorentz-type BG transformation is designed, which not only includes translation and rotation (which is somewhat similar to the affine transformation under the χ scheme), but also includes local adjustment of the data point position (which is somewhat similar to the local nonlinear transformation under the χ scheme) to maintain the "invariant" between the BIM model and the GIS data; the optimization scheme, the constructed algorithm should keep the correspondence between the BIM model and the GIS data (i.e., "invariant") unchanged after the BG transformation. Here, we can introduce a measurement function or energy function similar to the χ scheme, but add the concept of "space-time curvature" to measure the change in the correspondence before and after the transformation; the measurement function should contain two parts: one is the change in "space-time curvature" caused by the transformation, and the other is the deviation of the correspondence between the BIM model points and the GIS data points. In terms of optimization, the gradient descent method can be used to minimize the measurement function to find the best BG transformation parameters. During the optimization process, we can consider minimizing the change in "space-time curvature" while keeping the correspondence between the BIM model and GIS data "unchanged", so as to achieve the best alignment of the two data sets.

[0056] The ψ scheme is based on the idea of ​​geometric data. The basic technical idea is to use differential geometry and manifold learning methods to align data to achieve the integration of BIM models and GIS data. A basic data framework has been constructed, but there are some relatively advanced mathematical problems in its application.

[0057] The basic data framework is as follows: consider the geometric structure of the data manifolds in the two data systems and find the best alignment transformation. Corresponding to the two data systems of BIM and GIS, first set two manifolds M and N, represented by data sets X and Y respectively. Then the algorithm points to find a mapping T: M → N, so that T minimizes the Gromov-Wasserstein distance between the two manifolds. And the Gromov-Wasserstein distance is used to measure the similarity between the two metric spaces (M, d_M) and (N, d_N), where d_M and d_N are the distance functions on manifolds M and N respectively. The Gromov-Wasserstein distance is defined as: GW(X, Y, T) = inf_π∈Π(μ, ν) ∫_M×NL(d_M(x, x'), d_N(T(x), T(x'))) dπ(x, y) dπ(x', y'); where Π(μ, ν) is the set of all possible joint distributions π, whose marginal distributions are μ and ν, respectively, and μ and ν are probability distributions on M and N. The function L is a loss function, usually chosen as the square of the difference in distance. In terms of optimization, the goal is expressed as: minimize T GW(X_MDS, Y_MDS, T); to this end, we first perform multidimensional scaling analysis of the manifold. For data sets X and Y, we first apply multidimensional scaling analysis (MDS) to obtain their low-dimensional representations X_MDS and Y_MDS; then, calculate the distance matrix, and calculate the distance matrices D_M and D_N between point pairs in X_MDS and Y_MDS; then initialize the mapping T, based on a heuristic method; further use the EM algorithm or Sinkhorn iteration to update the mapping T and the joint distribution π, and update π in each iteration to minimize the Gromov-Wasserstein distance, and adjust T accordingly; set the iteration termination condition, such as stopping when the change in the Gromov-Wasserstein distance is less than a certain threshold or when the maximum number of iterations is reached; finally output the optimal mapping T that maps X_MDS to Y_MDS to achieve alignment of the two manifolds. With further support from mathematical knowledge, this model may have better development prospects than the current χ scheme.

[0058] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. The geometric transformation and alignment method between BIM model and GIS data in the power grid digital twin model is used for the secondary development of BIM model in the construction of power grid digital twin model, which is characterized by: The transformation alignment between BIM model and GIS data is based on linear affine transformation to construct the underlying algorithm, and multi-parameterized local nonlinear optimization matching is performed to obtain the main alignment algorithm framework; And as needed, the optimal dynamic transformation alignment between BIM model and GIS data is achieved through subsequent iterative optimization based on the model of the main alignment algorithm framework; The main alignment algorithm framework and its iterative optimization algorithm route are as follows: firstly, feature points are extracted and matched; feature points are extracted from BIM models and GIS data; the feature points are constructed to point to corners, edges or other significant geometric or geographical markers; Find the corresponding feature points between the BIM model and the GIS data, match the feature points in the BIM model and the GIS data based on the feature descriptors that mathematically describe the area around the feature points, where the feature descriptors are constrained based on rotation and scale invariance; compare the similarity of the feature descriptors to find the best matching feature point pair and optimize; further perform preliminary affine transformation estimation; the basic algorithm structure of the linear affine transformation is set to include two sets of parameters: coefficient parameters and translation parameters; based on this basic data structure, the first step is to introduce the least squares method to construct a linear equation group for estimating the affine transformation parameters and obtain the initial binary parameters by solving the equation group through elimination; Then, local nonlinear transformation parameters are estimated; the above-mentioned affine transformation results are used as input, and the geometric data deformation analysis tool is introduced into the main alignment algorithm framework to construct a measurement function containing data deformation energy and data terms. The linear data system is correspondingly derived and solved to find the transformation parameters that minimize the measurement function, and the local nonlinear transformation parameters are obtained and subsequently optimized as needed; The main alignment algorithm framework and the specific data process of its iterative optimization include: Step 1: Feature point extraction and matching; Existing computer vision algorithms are used to extract feature points from BIM models and GIS data. The following algorithmic data process is constructed for the matching of feature points: In order to achieve accurate matching of feature points between BIM models and GIS data, key points and their feature descriptions are first extracted based on the scale-invariant feature transformation algorithm. Key points are detected at different pre-specified scales or global scales and a feature descriptor is generated for each key point. The descriptor is consistent and invariant to rotation and scaling, so that it can be used to match between different data and their geometric representations; Next, the feature descriptors extracted from the BIM model are matched with the feature descriptors in the GIS data. The data distance is used as a measure of the difference between the two feature descriptors. The data square sum or square root distance between the feature descriptors in each BIM model and all feature descriptors in the GIS data is calculated, and it is matched with the nearest corresponding data. Figuratively speaking, this data process is to find the most matching data partner in the GIS data for each key point in the BIM model; The next step is to carry out subsequent directional matching optimization to improve the accuracy and robustness of matching. Specifically, two optimization strategies are constructed: α, a ratio test to exclude unreliable matches. On the basis of the distance between the feature descriptor in each BIM model and the nearest neighbor in the GIS data, the distance to the next nearest neighbor is added. If the distance ratio between the nearest neighbor and the next nearest neighbor is less than a preset threshold, that is, a value less than 1, the match is considered valid, thereby excluding ambiguous matches and improving the reliability of matching; β, geometric consistency test. This test process is constructed to improve the consistency of the geometric relationship of data between matching points. The data process first estimates the initial transformation model, that is, the affine transformation adopted, and verifies the proportion of all matching points that conform to the transformation model by randomly selecting a small number of matching points. On this basis, the largest set of matching points that conform to it is obtained through repeated iterations; Step 2: Preliminary affine transformation estimation; The basic algorithm structure of the linear affine transformation is set to include two sets of parameters: coefficient parameters and translation parameters; Specifically, the basic algorithm structure of the linear affine transformation is set to: X' = AX + b; Where X is the original coordinate, A is a 3 × 3 transformation matrix, b is the translation vector, and X' is the transformed coordinate; Based on this basic data structure, the first step is to introduce the least squares method to construct a linear equation group for estimating the affine transformation parameters A and b, and obtain the initial binary parameters by eliminating the equation group; Further, the optimization process of the preliminary affine transformation estimation is constructed. Since the preliminary affine transformation estimation is a linear transformation, the existing RANSAC algorithm is directly used here to eliminate incorrect matches to optimize the transformation parameters; Step 3: Estimation of local nonlinear transformation parameters; Specifically, the geometric data deformation analysis tool is introduced into the main alignment algorithm framework to optimize the local nonlinear transformation parameters. Both BIM and GIS data naturally have geometric spatial attributes, and the local nonlinearity of the data can be vividly imported and corresponded with the geometric shape transformation and data alignment; the data processing of geometric shape transformation and data alignment finds the best deformation mapping by minimizing a measurement function; at this time, firstly, a measurement function E(f) containing data deformation energy and data terms is constructed according to the following data standards. This function consists of two parts: one part is the deformation energy, which is used to punish excessive deformation; the other part is the data term, which ensures that the transformed point is as close to the target point as possible; based on this measurement function, the smoothness of the data deformation of the two system components of BIM and GIS and the correspondence between the data points are balanced; on this basis, a transformation function for characterizing the data transformation process is constructed according to the following standards The transformation function is set to include two sub-items: the linear transformation sub-item corresponds to the data representation of the rotation and / or scaling and / or translation operation, and the nonlinear deformation sub-item corresponds to the data representation of the local geometric adjustment, thereby finely adjusting the local details while maintaining the overall structure; at this time, in order to find the optimal transformation parameters based on the linear affine transformation, the problem of minimizing the measure function is transformed into a linear system problem. Specifically, the data processing process is set to set the partial derivative of the measure function with respect to each parameter to zero, thereby obtaining a set of linear equations, and this linear equation group is associated with the parameters including the linear transformation sub-item and the nonlinear deformation sub-item; in this way, after obtaining the linear system represented by the linear equation group, the numerical method is directly embedded for solving, such as embedding LU decomposition or conjugate gradient method to solve the offline system to obtain a set of optimal parameters so that the measure function reaches the minimum value and the corresponding optimal deformation mapping; and further construct the subsequent optimization data process as needed.

2. The geometric transformation and alignment method between the BIM model and the GIS data in the power grid digital twin model according to claim 1 is characterized in that: In step 3, the subsequent data optimization process includes: the optimization strategy of local nonlinear transformation is constructed as an application feedback model, specifically, the above-mentioned local nonlinear transformation function is applied to any point in the BIM and / or GIS source data space using the solved parameters, and its position in the GIS and / or BIM target data space is calculated, and the linear and nonlinear alignment of the source data to the target data, the proximity of the transformed point to the target point, and the smoothness of the data deformation are monitored and compared; the feedback data model is iterated and optimized according to the monitoring and comparison results.

3. A fully perceptual self-optimizing power grid project digital twin model system based on the geometric transformation and alignment method of the BIM model and GIS data in the power grid digital twin model of claim 1, characterized in that: The model system is further developed based on the BIM model technology for multi-system data fusion. When establishing the BIM model, geometric modeling and computer graphics technology are used to create an accurate geometric three-dimensional model. At the same time, based on the geometric transformation and alignment method of the BIM model and GIS data, it is integrated with the power grid GIS to simulate the full life cycle data of the power construction project to build a consistent coordination and comprehensive perception of the power construction data. The key steps in constructing the model system include: Data specification and standardization: formulate unified data specifications and data standards for power construction projects to ensure data consistency and interoperability; determine data classification, coding, format and interface standards to facilitate data integration and sharing; BIM model integration construction and digital twin platform development: Use BIM technology to create a three-dimensional model of the power construction project, including buildings, equipment, and pipelines; integrate the BIM model with the power grid GIS geographic information system and / or the power grid monitoring and data acquisition system and / or other necessary systems to achieve the integration of spatial data and real-time monitoring data; Among them, the integration of the BIM model and the power grid GIS geographic information system is based on the geometric transformation and alignment method of the BIM model and GIS data; Then, based on the above integrated BIM model, a digital twin platform that supports multi-dimensional data integration and real-time data processing is constructed, and supports real-time update and dynamic visualization of data; Full-chain data integration: Integrate data from all aspects of power construction project planning, design, construction, and operation and maintenance, including but not limited to drawings, plans, progress, costs, quality, and safety information, and realize real-time flow and update of data; Development of self-optimizing algorithms: Based on data analysis and machine learning, self-optimizing algorithms are built to predict and optimize the performance and cost of power construction projects. The algorithms automatically adjust the operating parameters of power construction projects based on historical and real-time data. Establishment of a digital collaborative ecosystem: Establish an internal and external digital collaborative ecosystem based on power construction project participants, suppliers, and regulatory agencies, and achieve data sharing and collaboration through application programming interfaces (APIs) and data sharing protocols; Data sharing and collaborative support: Build a set of data sharing mechanisms to coordinate the access and use of required data by power construction project participants and support collaborative interactions in the construction of power grid power projects, including design collaboration, construction collaboration, and operation and maintenance collaboration.

4. The digital twin model system of the fully-sensing self-optimizing power grid project according to claim 3 is characterized by: For data specification and standardization steps, statistical tools are introduced to analyze the characteristics of data and establish a set of unified data standards, including standards for data collection, processing, storage and exchange; For the development steps of the self-optimization algorithm, the development of the self-optimization algorithm is based on the technology of machine learning and artificial intelligence, using supervised learning and unsupervised learning methods to predict and optimize the performance of power construction projects; considering building a dedicated neural network to establish a complex local nonlinear model and using a genetic algorithm to find the optimal solution; For the steps of establishing a digital collaborative ecosystem, a system analysis is conducted based on the architectural design of the information system and network theory is used to optimize the process of data sharing and collaborative work. For data sharing and collaborative support, encryption algorithms, data compression technologies or network communication protocols are introduced or constructed with the goal of ensuring that data can be securely shared between different systems and organizations.

5. The digital twin model system of the fully-sensing self-optimizing power grid project according to claim 3 is characterized by: The model system also includes the following optional subsequent construction steps: secondary customization of multiple business scenarios, customizing the application scenarios of digital twin models including asset management, risk assessment, and fault diagnosis according to different business needs; and providing flexible user interfaces and reporting tools through data multi-channel ports; The model system also includes the following subsequent optional construction steps: security and privacy protection, considering data security and privacy protection in system design, and taking encryption and access control measures; at the same time, self-screening of data legalization with manual supervision or manual intervention based on the pre-imported legal and regulatory database.