A transformer digital twin model lightweight method and system

Abstract

The application provides a transformer digital twin model lightweight method and system, and relates to the field of power equipment digital twin. The method comprises the following steps: obtaining an initial three-dimensional model; obtaining model feedback requirements, and generating a model key feature data set based on the model feedback requirements; constructing a model simplification scheme or a lightweight strategy according to the model key feature data set and the initial three-dimensional model; obtaining real-time monitoring data of the transformer based on real-time monitoring parameter sets corresponding to the transformer three-dimensional model in the model simplification scheme; and substituting the real-time monitoring data of the transformer into the model simplification module, selectively adjusting the model simplification scheme, or following the lightweight strategy to realize the lightweight of the transformer three-dimensional model. The application can realize the lightweight of the power transformer three-dimensional model, ensure the accuracy of the real-time data reflected by the transformer three-dimensional model, reduce the model data volume and improve the response speed of the system.

Description

Technical Field

[0001] This disclosure relates to the field of digital twin technology for power equipment, and in particular to a lightweight method and system for digital twin models of transformers. Background Technology

[0002] With the development of digital twin technology, building 3D models for real-time monitoring, fault prediction, and operation and maintenance management of transformers has become a common method. The 3D digital twin model needs to accurately reflect the actual state of the transformer, including its structure, temperature distribution, vibration information, etc. Due to the complexity of the transformer structure and the high requirements for real-time interaction in digital twins, there is a problem of excessive computational data volume, placing high demands on computing and storage resources, making it difficult to meet the requirements of efficiency and real-time performance.

[0003] Existing 3D models often involve large amounts of data and complex details, placing high demands on storage and computing resources. In particular, in edge computing devices and real-time monitoring scenarios, existing models often struggle to meet the requirements for efficiency and real-time performance.

[0004] Therefore, there is an urgent need for a lightweighting technique for 3D models that can reduce the amount of model data and improve the system's response speed while ensuring model accuracy. This is a lightweighting method and system for 3D models of digital twins of transformers. Summary of the Invention

[0005] This disclosure provides a lightweight method and system for transformer digital twin models, which solves the technical problem that existing 3D models often have large data volume and complex details, and have high requirements for storage and computing resources. Especially in edge computing devices and real-time monitoring scenarios, existing models often cannot meet the requirements of efficiency and real-time performance.

[0006] According to a first aspect of this disclosure, a lightweight method for a transformer digital twin model is provided. The method includes: acquiring an initial three-dimensional model; acquiring model feedback requirements and generating a key feature dataset based on these requirements; constructing a model simplification scheme or a lightweighting strategy based on the key feature dataset and the initial three-dimensional model; wherein, constructing the model simplification scheme based on the key feature dataset and the initial three-dimensional model includes: substituting the key feature dataset and the initial three-dimensional model into a model simplification module to obtain a model simplification scheme; acquiring real-time monitoring data of the transformer based on the real-time monitoring parameter set corresponding to the transformer three-dimensional model in the model simplification scheme; substituting the real-time monitoring data of the transformer into the model simplification module, selectively adjusting the model simplification scheme, or following the lightweighting strategy to achieve lightweighting of the transformer three-dimensional model.

[0007] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein obtaining model feedback requirements and generating a model key feature dataset based on the model feedback requirements includes: obtaining model feedback requirements, wherein the model feedback requirements include at least one monitoring requirement from the user for at least one of the transformer 3D models and the priority level of each monitoring requirement; and determining at least one necessary key feature of the model and the priority level of each necessary key feature based on the model feedback requirements.

[0008] Based on at least one essential key feature of the model and the priority level of each essential key feature, non-essential key features corresponding to the essential key features of the model and their priority levels are selectively generated; the at least one essential key feature of the model, the priority level of each essential key feature, the selectively generated non-essential key features corresponding to the essential key features of the model and their priority levels are filtered to obtain at least one filtered essential key feature of the model and non-essential key features corresponding to the essential key features of the model, or at least one filtered essential key feature of the model; a key feature dataset of the model is constructed based on the at least one filtered essential key feature of the model and the non-essential key features corresponding to the essential key features of the model, or at least one filtered essential key feature of the model.

[0009] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the step of filtering at least one essential model key feature, the priority level of each essential model key feature, selectively generated non-essential model key features corresponding to the essential model key features, and the priority levels of the non-essential model key features yields at least one filtered essential model key feature and corresponding non-essential model key features. Alternatively, the filtered essential model key feature includes: when no corresponding non-essential model key features and their priority levels are generated, the essential model key features are selectively and initially eliminated based on their priority levels, yielding... The process involves obtaining multiple sets of essential key features for the model after initial elimination, where each set includes at least several essential key features. Based on each set of essential key features after initial elimination, the initial 3D model is simplified and simulated to obtain simplified initial 3D models. Based on each simplified initial 3D model, the corresponding dataset is determined. Based on the corresponding dataset, the multiple sets of essential key features after initial elimination are filtered to obtain multiple sets of essential key features. Based on the multiple sets of essential key features, multiple essential key features from one set are selected as at least one essential key feature of the model after filtering.

[0010] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the step of filtering at least one essential key feature of the model, the priority level of each essential key feature of the model, selectively generated non-essential key features of the model corresponding to the essential key feature of the model, and the priority level of the non-essential key features of the model, to obtain at least one essential key feature of the model and non-essential key features of the model corresponding to the essential key feature of the model, or, the at least one essential key feature of the model after filtering further includes: when non-essential key features of the model corresponding to the essential key feature of the model and the priority level of the non-essential key features of the model are generated: based on the priority level of each essential key feature of the model, the essential key features of the model are selectively initially eliminated to obtain multiple sets of essential key features of the model after initial elimination, wherein each set of essential key features of the model includes at least a plurality of essential key features of the model;

[0011] Based on the priority of each model's non-essential key features and multiple sets of model essential key features after initial elimination, at least one model non-essential key feature corresponding to the model essential key feature set is selectively generated to constitute a model key feature set. Each set of model key feature sets includes at least one set of model essential key features and at least one model non-essential key feature corresponding to the model essential key feature set. Based on each set of model key feature sets and each set of model essential key feature sets, the initial 3D model is simplified and simulated to obtain simplified initial 3D models. Based on each simplified initial 3D model, the corresponding dataset is determined. Based on each simplified initial 3D model, the dataset corresponding to each simplified initial 3D model is determined. The dataset corresponding to the dimensional model is used to filter multiple sets of essential key features and multiple sets of key features of the model after initial elimination, resulting in multiple sets of essential key features and / or multiple sets of key features of the model. Based on the multiple sets of essential key features and / or multiple sets of key features of the model, multiple essential key features from one set of essential key features are selected as at least one essential key feature of the model after filtering. Alternatively, at least one essential key feature from one set of key features of the model and the corresponding non-essential key feature of the model are selected as at least one essential key feature of the model after filtering and the corresponding non-essential key feature of the model after filtering.

[0012] In addition to the aspects described above and any possible implementation, a further implementation is provided, wherein the selective preliminary elimination of the essential key features of the model based on the priority level of each essential key feature of the model, to obtain multiple sets of essential key features of the model after preliminary elimination, includes: determining the necessary priority level threshold corresponding to the essential key features of the model based on model feedback requirements; if the priority level of the essential key feature of the model is lower than the necessary priority level threshold, then the essential key feature of the model is screened out; otherwise, the essential key feature of the model is retained; to obtain all retained essential key features of the model; and randomly matching all retained essential key features of the model to form multiple sets of essential key features of the model after preliminary elimination.

[0013] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein substituting the model key feature dataset and the initial 3D model into the model simplification module to obtain the model simplification scheme includes: generating a simplification process corresponding to the model key feature dataset based on the model key feature dataset, wherein the simplification process includes at least multiple simplification types, the step of each simplification type, and simplification parameters for each simplification type; simplifying the initial 3D model based on the model key feature dataset and the simplification process corresponding to the model key feature dataset to generate a preliminary simplified transformer 3D model; simulating the preliminary simplified transformer 3D model to obtain simulation data of the preliminary simplified transformer 3D model; selectively iteratively optimizing the preliminary simplified transformer 3D model based on the simulation data of the preliminary simplified transformer 3D model to obtain an iteratively optimized transformer 3D model, and using the iteratively optimized transformer 3D model as the simplified transformer 3D model; generating a real-time monitoring parameter set corresponding to the transformer 3D model based on the simplified transformer 3D model and the model key feature dataset; thereby constituting the model simplification scheme and outputting it.

[0014] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the step of generating a simplified process corresponding to the model key feature dataset based on the model key feature dataset includes: selectively determining multiple simplification types based on whether the model key feature dataset contains non-essential key features corresponding to the necessary key features of the model; generating multiple sets of primary simplification processes based on the multiple simplification types and the model key feature dataset, including: combining the multiple simplification types to form multiple sets of simplification type combinations; generating multiple sets of simplified parameter combinations corresponding to each set of simplified type combinations, wherein the multiple sets of simplified parameter combinations include simplified parameters for each simplification type; and generating multiple sets of simplified parameter combinations based on the multiple sets of simplified type combinations. Each set of simplified type combinations and the corresponding multiple sets of simplified parameter combinations constitute multiple sets of primary simplification processes. Based on the multiple sets of primary simplification processes, the initial 3D model is simplified and simulated respectively to obtain each simplified initial 3D model. Based on each simplified initial 3D model, the multiple sets of primary simplification processes are filtered to obtain filtered primary simplification processes. Based on the initial 3D model corresponding to each filtered primary simplification process, the corresponding dataset is determined. Based on the corresponding dataset of the initial 3D model corresponding to each filtered primary simplification process, a set of filtered primary simplification processes is selected as the simplification process corresponding to the key feature dataset of the model.

[0015] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein a model simplification scheme or a lightweight strategy is constructed based on the model key feature dataset and the initial 3D model; the lightweight strategy is followed to achieve lightweighting of the transformer geometric-physical digital twin model, including: extracting features and labeling regions from the key information of the geometric model in the model key feature dataset to obtain feature data; calculating the vertex saliency of the geometric model using a vertex saliency-driven classification edge shrinkage algorithm; classifying all edges in the feature data according to the saliency of adjacent vertices into classes I, II, and III; constructing an edge shrinkage cost function with saliency penalty; prioritizing the shrinkage of class I edges using a min-heap priority queue according to the edge shrinkage cost function, synchronously updating the topology and saliency data until the number of faces ≤ the target number or the shrinkage cost ≥ the upper limit of error is satisfied; performing mesh smoothing on the lightweight model, and ensuring that HD(A, B) ≤ 3% through Hausdorff distance verification, otherwise reconstructing the cost function; and completing the lightweighting of the transformer 3D model.

[0016] In addition to the aspects described above and any possible implementation, a further implementation is provided, wherein generating the real-time monitoring parameter set corresponding to the three-dimensional transformer model based on the simplified three-dimensional transformer model and the model key feature dataset includes: determining at least one necessary monitoring feature and the corresponding unnecessary monitoring feature in the real-time monitoring parameter set based on at least one necessary key feature of the model in the model key feature dataset and the corresponding unnecessary key feature of the model, or at least one necessary key feature of the model; determining an update threshold for each necessary monitoring feature in the real-time monitoring parameter set and the update threshold for each necessary monitoring feature, or the update threshold for each necessary monitoring feature, based on the simplified three-dimensional transformer model; and constructing the real-time monitoring parameter set corresponding to the three-dimensional transformer model using at least one necessary monitoring feature in the real-time monitoring parameter set and the corresponding unnecessary monitoring feature, or at least one necessary monitoring feature, and the update threshold corresponding to each monitoring feature.

[0017] According to a second aspect of this disclosure, a lightweight system for a digital twin 3D model of a transformer is provided. The system includes: a communication device, a data acquisition device, a model simplification module, and a control device. The communication device is used to realize information interaction between the data acquisition device and the transformer and a management terminal. The data acquisition device is at least used to acquire real-time monitoring data of the transformer. The model simplification module is used to generate a model simplification scheme based on a dataset of key model features and an initial 3D model, and selectively adjusts the model simplification scheme based on the real-time monitoring data of the transformer to simplify the 3D model of the transformer. The control device includes a memory and a processor. The memory stores a computer program, and the processor executes the program to implement the methods described above.

[0018] The above-disclosed technical solutions have at least one or more of the following beneficial effects:

[0019] By obtaining model feedback requirements, a key feature dataset of the model is generated, providing the data foundation required for the model simplification module. By substituting the initial 3D model and the key feature dataset into the model simplification module, a model simplification scheme is obtained, enabling automatic generation of the simplification scheme. This ensures that while simplifying the transformer's 3D model, the simplification scheme also guarantees real-time monitoring of the transformer and ensures that the real-time monitoring parameter set matches the transformer's 3D model. Based on the real-time monitoring parameter set corresponding to the transformer's 3D model in the simplification scheme, real-time monitoring data of the transformer is obtained. This real-time monitoring data is then substituted into the model simplification module, selectively adjusting the model simplification scheme to further simplify the transformer's 3D model. This ensures that while optimizing the simplified 3D model, the accuracy of the 3D model in reflecting real-time data is maintained. This allows for continuous adjustment and optimization of lightweight processing parameters based on actual usage, achieving a balance between lightweight design and accuracy. It avoids the technical problem that existing 3D models often have large data volumes and complex details, requiring high storage and computing resources, especially in edge computing devices and real-time monitoring scenarios, where existing models often fail to meet the demands for efficiency and real-time performance.

[0020] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0021] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0022] Figure 1 A flowchart of a method for lightweighting a digital twin 3D model of a transformer according to an embodiment of the present disclosure is shown;

[0023] Figure 2 A schematic diagram of the main process of a lightweight method for a digital twin 3D model of a transformer according to Embodiment 2 of this disclosure is shown.

[0024] Figure 3 A detailed flowchart of a method for lightweighting a digital twin 3D model of a transformer according to Embodiment 2 of this disclosure is shown.

[0025] Figure 4 A block diagram of a lightweight system for a digital twin 3D model of a transformer is shown according to an embodiment of the present disclosure.

[0026] List of reference numerals in the attached diagram:

[0027] 200: Lightweight system; 201: Control device; 2011: Processor; 2012: Memory; 2013: Program code; 202: Model simplification module; 203: Communication device; 204: Data acquisition device. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0029] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0030] Example 1:

[0031] See appendix Figure 1 , Figure 1A flowchart illustrating a lightweight method for a transformer digital twin 3D model according to Embodiment 1 of this disclosure is shown. Figure 1 As shown, the lightweight method for a transformer digital twin 3D model in Embodiment 1 of this disclosure mainly includes the following steps S101-S105.

[0032] Step S101: Obtain the initial 3D model;

[0033] Specifically, obtaining the initial 3D model includes:

[0034] Obtain the initial dataset of the transformer, wherein the initial dataset of the transformer includes at least the internal structure data and the external structure data of the transformer;

[0035] Based on the initial dataset of transformers, an initial three-dimensional model of the transformer is constructed.

[0036] Specifically, the initial three-dimensional model of the transformer can be constructed using high-precision three-dimensional modeling tools in the prior art. The choice of construction method here is only an example. In actual testing, those skilled in the art can choose according to actual needs, as long as it can realize the construction of the initial three-dimensional model of the transformer based on the initial dataset of the transformer. It will not be elaborated here.

[0037] Step S102: Obtain model feedback requirements and generate a dataset of key model features based on these requirements;

[0038] Specifically, the step of obtaining model feedback requirements and generating a dataset of key model features based on those requirements includes:

[0039] Obtain model feedback requirements, wherein the model feedback requirements include at least one monitoring requirement from the user for at least one of the transformer 3D models and the priority level of each monitoring requirement;

[0040] Based on the model feedback requirements, determine at least one essential key feature of the model and the priority level of each essential key feature of the model.

[0041] Based on at least one essential key feature of the model and the priority level of each essential key feature, non-essential key features of the model corresponding to the essential key features of the model and the priority level of the non-essential key features of the model are selectively generated.

[0042] The model essential key features, the priority level of each model essential key feature, the selectively generated non-essential key features corresponding to the model essential key features, and the priority level of the model non-essential key features are filtered to obtain at least one model essential key feature and a non-essential key feature corresponding to the model essential key feature, or at least one model essential key feature after filtering.

[0043] A model key feature dataset is constructed based on at least one of the selected essential key features of the model and the corresponding non-essential key features of the model, or on at least one of the selected essential key features of the model.

[0044] Specifically, the monitoring requirement of the at least one can be any type of parameter of key components such as transformer casing, winding, core, oil tank, and cooling system, and the priority level of each monitoring requirement can be set by the technicians themselves.

[0045] Specifically, the selective generation of non-essential model features corresponding to the essential model features and their priority levels, based on at least one essential model feature and the priority level of each essential model feature, includes:

[0046] If the essential key features of the model cannot be determined by other parameters, then it is determined that the essential key features of the model do not require non-essential key features.

[0047] Otherwise, if it is determined that the essential key features of the model require non-essential key features, then non-essential key features of the model corresponding to the essential key features of the model are generated, and the priority level of the non-essential key features of the model is determined based on the priority level of the essential key features of the model.

[0048] Specifically, the step of filtering at least one essential key feature of the model, the priority level of each essential key feature, selectively generated non-essential key features corresponding to the essential key features, and the priority level of the non-essential key features yields at least one essential key feature of the model and non-essential key features corresponding to the essential key features; or, the at least one essential key feature of the model after filtering includes:

[0049] When no non-essential key features corresponding to the essential key features of the model and no priority ranking of the non-essential key features of the model are generated:

[0050] Based on the priority level of each essential key feature of the model, the essential key features of the model are selectively eliminated in the preliminary stage, resulting in multiple sets of essential key features of the model after preliminary elimination. Each set of essential key features of the model includes at least a number of essential key features of the model, and the multiple essential key features of the model in each set of essential key features of the model are not completely consistent.

[0051] Based on the set of essential key features of each model after initial elimination, the initial three-dimensional models are simplified and simulated to obtain simplified initial three-dimensional models.

[0052] Based on each simplified initial 3D model, determine the corresponding dataset for each simplified initial 3D model;

[0053] Based on the dataset corresponding to each simplified initial 3D model, the necessary key feature sets of multiple models after preliminary elimination are filtered to obtain the filtered sets of necessary key feature sets of multiple models.

[0054] Based on the multiple sets of essential key features of the model after screening, select multiple essential key features from one set of essential key features of the model as at least one essential key feature of the model after screening.

[0055] In some embodiments, specifically, the filtering of multiple sets of essential key features of the models after preliminary elimination, based on the dataset corresponding to each simplified initial 3D model, to obtain multiple sets of essential key features of the models after filtering includes:

[0056] If the simplified initial 3D model corresponding to any set of essential key features of the model after the initial elimination matches all the essential key features of the model in the set of essential key features of the model, then the set of essential key features of the model is retained.

[0057] If the dataset corresponding to the simplified initial 3D model of any set of essential key features of the model after initial elimination does not match all essential key features of the model in the dataset of essential key features of the model, and the number of unmatched essential key features of the model reaches a preset threshold, then the set of essential key features of the model is eliminated.

[0058] This is to obtain a set of essential key features for multiple models after filtering.

[0059] Alternatively, in some embodiments, specifically, the step of filtering the multiple sets of essential key features of the models after preliminary elimination based on the dataset corresponding to each simplified initial 3D model, to obtain the filtered multiple sets of essential key features of the models, includes:

[0060] If the dataset corresponding to the simplified initial 3D model of any set of essential key features of the model after preliminary elimination cannot meet the model feedback requirements, then the set of essential key features of the model after preliminary elimination is determined to be unacceptable, and the set of essential key features of the model after preliminary elimination is eliminated.

[0061] Otherwise, if the set of essential key features of the model after the initial elimination is determined to meet the requirements, the set of essential key features of the model after the initial elimination will be retained.

[0062] This is to obtain a set of essential key features for multiple models after filtering.

[0063] Specifically, the step of selecting multiple essential key features from a set of filtered essential key features as at least one essential key feature after filtering includes:

[0064] Based on the filtered sets of multiple sets of essential key features for the model, the matching degree of each set of essential key features for the model to all essential key features in the dataset of essential key features for the model is obtained.

[0065] Based on the selected sets of essential key features for multiple models, the simplification degree of each set of essential key features for models is determined.

[0066] Based on the simplification of each set of essential key features of the model and the matching degree of each set of essential key features of the model to all essential key features in the model key feature dataset, a comprehensive evaluation score is obtained for each set of essential key features of the model.

[0067] Select multiple essential key features from the set of essential key features with the highest comprehensive evaluation score, and use them as at least one essential key feature of the model after screening.

[0068] Specifically, the step of filtering at least one essential key feature of the model, the priority level of each essential key feature, selectively generated non-essential key features corresponding to the essential key features, and the priority level of the non-essential key features yields at least one essential key feature of the model and non-essential key features corresponding to the essential key features; alternatively, the at least one essential key feature of the model further includes:

[0069] When generating non-essential key features of the model corresponding to the essential key features of the model, and the priority levels of the non-essential key features of the model:

[0070] Based on the priority level of each essential key feature of the model, the essential key features of the model are selectively eliminated in the preliminary stage, resulting in multiple sets of essential key features of the model after preliminary elimination. Each set of essential key features of the model includes at least a number of essential key features of the model, and the multiple essential key features of the model in each set of essential key features of the model are not completely consistent.

[0071] Based on the priority of each model's non-essential key features and the multiple sets of model essential key features after initial elimination, at least one model non-essential key feature corresponding to the set of model essential key features is selectively generated to form a set of model key features. Each set of model key features includes at least one set of model essential key features and at least one model non-essential key feature corresponding to the set of model essential key features.

[0072] Based on the key feature set of each model and the necessary key feature set of each model, the initial three-dimensional model is simplified and simulated to obtain the simplified initial three-dimensional model.

[0073] Based on each simplified initial 3D model, determine the corresponding dataset for each simplified initial 3D model;

[0074] Based on the dataset corresponding to each simplified initial 3D model, the necessary key feature sets of multiple models and the key feature sets of multiple models after preliminary elimination are filtered to obtain the filtered sets of necessary key feature sets of multiple models and / or sets of key feature sets of multiple models.

[0075] Based on the filtered sets of multiple sets of essential key features and / or multiple sets of key features, select multiple essential key features from one set of essential key features as at least one essential key feature of the filtered model; or, select at least one essential key feature from one set of key features and the corresponding non-essential key features of the model as at least one essential key feature of the filtered model and the corresponding non-essential key features of the model.

[0076] Specifically, based on the priority level of each essential key feature of the model, the essential key features of the model are selectively initially eliminated, resulting in multiple sets of essential key features of the model after initial elimination, including:

[0077] Based on the model feedback requirements, determine the necessary priority level thresholds corresponding to the necessary key features of the model.

[0078] If the priority level of the essential key features of the model is lower than the essential priority level threshold, then the essential key features of the model are screened out.

[0079] Otherwise, the essential key features of the model are retained;

[0080] To obtain all the essential key features of the retained model;

[0081] All retained essential key features of the model are randomly matched to form multiple sets of essential key features of the model after initial elimination.

[0082] Specifically, determining the necessary priority thresholds corresponding to the essential key features of the model based on model feedback requirements includes:

[0083] If any of the monitoring requirements in the model feedback requirements can be reflected by multiple essential key features of the model, then the average priority level of the multiple essential key features of the model corresponding to each monitoring requirement shall be taken as one of the essential priority level values.

[0084] Based on all necessary priority level values, determine the necessary priority level thresholds corresponding to the necessary key features of the model;

[0085] Specifically, the method further includes:

[0086] If any of the monitoring requirements in the model feedback requirements can be reflected by at least one essential key feature of the model and at least one non-essential key feature of the model, then the non-essential key feature of the model corresponding to the monitoring requirement shall be taken as one of the non-essential priority level values.

[0087] Based on all the non-essential priority level values, determine the non-essential priority level thresholds corresponding to the non-essential key features of the model.

[0088] Specifically, determining the corresponding dataset for each simplified initial 3D model based on each simplified initial 3D model includes:

[0089] Based on each simplified initial 3D model, obtain the direct monitoring type and the indirect monitoring type of multiple monitoring data mapped by each simplified initial 3D model;

[0090] The simplified initial three-dimensional model is composed of the direct monitoring type of multiple monitoring data and the indirect monitoring type of multiple monitoring data.

[0091] In some embodiments, specifically, the step of filtering the multiple sets of necessary key features of the models and the multiple sets of key features of the models after preliminary elimination, based on the dataset corresponding to each simplified initial 3D model, to obtain the filtered multiple sets of necessary key features of the models and / or multiple sets of key features of the models includes:

[0092] If the simplified initial 3D model corresponding to any set of essential key features of the model after the initial elimination matches all the essential key features of the model in the set of essential key features of the model, then the set of essential key features of the model is retained.

[0093] If the dataset corresponding to the simplified initial 3D model of any set of essential key features of the model after initial elimination does not match all essential key features of the model in the dataset of essential key features of the model, and the number of unmatched essential key features of the model reaches a preset threshold, then the set of essential key features of the model is eliminated.

[0094] And / or,

[0095] If the dataset corresponding to the simplified initial 3D model corresponding to any set of model key features matches all the necessary key features of the model in the set of model key features, then the set of model key features is retained.

[0096] If the dataset corresponding to the simplified initial 3D model corresponding to any set of model key features does not match all the necessary key features of the model in the set of model key features, and the number of unmatched necessary key features of the model reaches a preset threshold, then the set of model key features will be removed.

[0097] To obtain the necessary key feature set of multiple sets of models and / or the key feature set of multiple sets of models after screening.

[0098] Alternatively, in some embodiments, specifically, the step of filtering the multiple sets of necessary key features of the models and the multiple sets of key features of the models after preliminary elimination, based on the dataset corresponding to each simplified initial 3D model, to obtain the filtered multiple sets of necessary key features of the models and / or multiple sets of key features of the models includes:

[0099] If the dataset corresponding to the simplified initial 3D model of any set of essential key features of the model after preliminary elimination cannot meet the model feedback requirements, then the set of essential key features of the model after preliminary elimination is determined to be unacceptable, and the set of essential key features of the model after preliminary elimination is eliminated.

[0100] If the dataset corresponding to the simplified initial 3D model corresponding to any set of key model features cannot meet the model feedback requirements, then the set of key model features is deemed unacceptable and is removed.

[0101] Otherwise, the set of essential key features of the model after the initial elimination is determined, or if the set of key features of the model meets the requirements, the set of essential key features of the model after the initial elimination, or the set of key features of the model, is retained.

[0102] To obtain the filtered set of necessary key features for multiple models and / or the set of key features for multiple models.

[0103] Specifically, the step of selecting multiple essential key features from a set of essential key features and / or multiple sets of key features after screening, as at least one essential key feature of the model after screening, or selecting at least one essential key feature from a set of key features and corresponding non-essential key features of the model after screening, as at least one essential key feature of the model and corresponding non-essential key features of the model after screening, includes:

[0104] Based on the filtered sets of multiple sets of essential key features for the model and / or multiple sets of key features for the model, the matching degree of each set of essential key features for the model and / or each set of key features for the model to all essential key features in the set of key features for the model is obtained.

[0105] Based on the selected sets of essential key features for multiple models and / or sets of key features for multiple models, determine the simplification of each set of essential key features for multiple models and / or sets of key features for multiple models.

[0106] Based on the selected sets of essential key features for multiple models and / or sets of key features for multiple models, the accuracy of each set of essential key features for multiple models and / or sets of key features for multiple models is obtained.

[0107] Based on the matching degree of each set of essential key features of the model and / or the matching degree of each set of essential key features of the model to all essential key features in the set of essential key features of the model, the simplification degree of each set of essential key features of the model and / or the accuracy of each set of essential key features of the model and / or the accuracy of each set of essential key features of the model and / or the accuracy of each set of essential key features of the model, a comprehensive evaluation score is obtained for each set of essential key features of the model and / or the matching degree of each set of essential key features of the model to all essential key features in the set of essential key features of the model.

[0108] If the set with the highest comprehensive evaluation score is the set of essential key features of the model, then select multiple essential key features from the set of essential key features of the model with the highest comprehensive evaluation score as at least one essential key feature of the model after screening.

[0109] or,

[0110] If the set with the highest comprehensive evaluation score is the set of key features of the model, then at least one essential key feature of the model and the corresponding non-essential key feature of the model from the set of key features of the model with the highest comprehensive evaluation score are selected as at least one essential key feature of the model and the corresponding non-essential key feature of the model after screening.

[0111] In the above embodiments, by using model feedback requirements, multiple essential key features of the model and their priority levels are determined. This enables the accurate identification of key components affecting equipment performance and safety, i.e., essential key features of the model, based on the monitoring requirements of the transformer. Then, by selectively generating non-essential key features of the model, the accuracy of key features affecting equipment performance and safety is improved. Furthermore, by filtering all key features, the accuracy of the model's key feature dataset for monitoring the transformer is improved. Moreover, by filtering and simplifying non-critical components or details with minor impact, the complexity of the model is reduced.

[0112] Step S103: Construct a model simplification scheme or a lightweight strategy based on the model key feature dataset and the initial 3D model. The model simplification scheme includes at least a simplified 3D transformer model and a set of real-time monitoring parameters corresponding to the 3D transformer model.

[0113] In one embodiment, constructing a model simplification scheme based on the model key feature dataset and the initial 3D model includes: substituting the model key feature dataset and the initial 3D model into the model simplification module to obtain the model simplification scheme, specifically including:

[0114] Based on the model's key feature dataset, a simplified process corresponding to the model's key feature dataset is generated. The simplified process includes at least multiple simplification types, the steps of each simplification type, and simplification parameters for each simplification type.

[0115] Based on the model key feature dataset and the simplification process corresponding to the model key feature dataset, the initial three-dimensional model is simplified to generate a preliminary simplified three-dimensional transformer model.

[0116] The simplified three-dimensional model of the transformer was simulated to obtain simulation data of the simplified three-dimensional model of the transformer.

[0117] Based on the simulation data of the preliminary simplified three-dimensional transformer model, the preliminary simplified three-dimensional transformer model is selectively iteratively optimized to obtain the iteratively optimized three-dimensional transformer model, and the iteratively optimized three-dimensional transformer model is used as the simplified three-dimensional transformer model.

[0118] Based on the simplified 3D model of the transformer and the dataset of key features of the model, a set of real-time monitoring parameters corresponding to the 3D model of the transformer is generated.

[0119] To construct a simplified model and output it.

[0120] Specifically, the simplified process for generating a dataset corresponding to the model's key features based on the model's key feature dataset includes:

[0121] Based on whether there are non-essential key features in the model's key feature dataset that correspond to the necessary key features of the model, multiple simplification types are selectively determined;

[0122] Based on multiple simplification types and the model's key feature dataset, multiple sets of primary simplification processes are generated. These primary simplification processes may differ in their simplification types, the steps they take, or the simplification parameters for each simplification type.

[0123] Based on multiple sets of initial simplification processes, the initial three-dimensional model is simplified and simulated to obtain each simplified initial three-dimensional model.

[0124] Based on each simplified initial 3D model, multiple sets of primary simplified processes are filtered to obtain the filtered primary simplified processes;

[0125] Based on the initial 3D model corresponding to each filtered primary simplified process, determine the corresponding dataset of the initial 3D model corresponding to each filtered primary simplified process;

[0126] Based on the dataset corresponding to the initial 3D model of each filtered primary simplified process, a set of filtered primary simplified processes is selected as the simplified process corresponding to the key feature dataset of the model.

[0127] Specifically, the selective determination of multiple simplification types based on whether the model's key feature dataset contains non-essential key features corresponding to the model's essential key features includes:

[0128] If the model key feature dataset contains non-essential key features corresponding to the essential key features of the model, then multiple simplification types are determined, wherein the multiple simplification types include at least several of data compression, LOD optimization, and geometric simplification;

[0129] or,

[0130] If there are no non-essential key features in the model's key feature dataset that correspond to the model's essential key features, then multiple simplification types are determined, wherein the multiple simplification types include at least data compression and LOD optimization.

[0131] Specifically, the geometric simplification can sequentially include polygon simplification, smoothing, and surface texture compression. Polygon simplification can be achieved by simplifying the high-density polygon mesh in the 3D model, reducing the number of triangular faces, thereby preserving the geometric features of key components and ensuring visual fidelity. Smoothing can be achieved by smoothing the simplified surface, removing unnecessary facial details and reducing the rendering burden of the 3D model. Surface texture compression can be achieved by processing the surface texture map of the 3D model using compression technology, reducing the image file size while ensuring the clarity and realism of the texture. The geometric simplification settings here are only illustrative examples. In actual testing, those skilled in the art can set them according to actual needs, as long as the initial 3D model can be simplified through geometric simplification. Further details are omitted here.

[0132] Specifically, the data compression can include model data compression and material simplification in sequence. The model data compression can use a data compression algorithm (here, a wavelet transform-based or vertex compression algorithm) to compress the files in the 3D model, thereby reducing the file size of the model. Material simplification can reduce the number of unnecessary material types in the 3D model. The simplification method can be to replace them with uniform or similar materials, thereby reducing the computational load of rendering the 3D model. The data compression settings here are only illustrative examples. In actual testing, those skilled in the art can set them according to actual needs. As long as the initial 3D model can be simplified by compressing the data of the 3D model, it will not be elaborated further here.

[0133] Specifically, the LOD optimization can include hierarchical display and dynamic loading. Hierarchical display can be achieved by using different model resolutions based on viewing distance or display resolution. That is, a high-precision 3D model is displayed when viewed at close range, while a low-precision 3D model is automatically switched when viewed at a distance, thereby improving rendering efficiency. Dynamic loading can be achieved by using a block loading technique, dividing the 3D model into multiple regions, and ensuring that only the part within the field of view is loaded. Other regions are dynamically loaded when the viewing angle changes, thereby reducing memory usage.

[0134] Specifically, the process of generating multiple sets of primary simplification steps based on multiple simplification types and key feature datasets of the model includes:

[0135] Based on multiple simplified types, multiple sets of simplified type combinations are formed, wherein the simplified types in the multiple sets of simplified type combinations are different, or the steps in which each simplified type is located are different;

[0136] Based on multiple sets of simplified type combinations, multiple sets of simplified parameter combinations are generated for each set of simplified type combinations. The multiple sets of simplified parameter combinations include simplified parameters for each simplified type, and the simplified parameters for each simplified type are different among the multiple sets of simplified parameter combinations.

[0137] Multiple sets of simplified type combinations and multiple sets of simplified parameter combinations corresponding to each simplified type combination constitute multiple sets of primary simplification processes.

[0138] Specifically, the simplification method for the initial 3D model can be achieved by using a trained neural network model or by using an intelligent learning model. The choice of simplification method here is only an example. In actual testing, those skilled in the art can choose according to actual needs, as long as it can achieve the simplification of the initial 3D model based on the model's key feature dataset and the simplification process corresponding to the model's key feature dataset, and generate a preliminary simplified transformer 3D model. Further details are omitted here.

[0139] Specifically, the simulation method for simulating the initially simplified three-dimensional transformer model and the simplified simulation method for simulating the initial three-dimensional model can employ existing simulation techniques, or a neural network model for simulation or simplified simulation. The selection of simulation method and simplified simulation method here is only an illustrative example. In actual testing, those skilled in the art can choose according to actual needs, as long as it can achieve the simulation of the initially simplified three-dimensional transformer model to obtain the simulation data of the initially simplified three-dimensional transformer model, and the simplified simulation of the initial three-dimensional model according to the primary simplification process to obtain each simplified initial three-dimensional model. Further details are omitted here.

[0140] Specifically, the generation of the real-time monitoring parameter set corresponding to the three-dimensional transformer model based on the simplified three-dimensional transformer model and the model's key feature dataset includes:

[0141] Based on at least one essential key feature of the model in the model key feature dataset and the non-essential key feature of the model corresponding to the essential key feature, or at least one essential key feature of the model, determine at least one essential monitoring feature in the real-time monitoring parameter set and the non-essential monitoring feature corresponding to the essential monitoring feature, or at least one essential monitoring feature.

[0142] Based on the simplified three-dimensional model of the transformer, the update threshold of each necessary monitoring feature in the real-time monitoring parameter set and the update threshold of each necessary monitoring feature, or the update threshold of each necessary monitoring feature, are determined.

[0143] The real-time monitoring parameter set corresponding to the three-dimensional model of the transformer is obtained by using at least one necessary monitoring feature from the real-time monitoring parameter set and the non-necessary monitoring feature corresponding to the necessary monitoring feature, or at least one necessary monitoring feature and the update threshold corresponding to each monitoring feature.

[0144] Specifically, the method for determining the update threshold corresponding to each monitoring feature can be simplified by using a trained neural network model, or by using an intelligent learning model to simplify the initial three-dimensional model. The choice of method here is only an example. In actual testing, those skilled in the art can choose according to actual needs, as long as it can achieve the determination of the update threshold of each necessary monitoring feature in the real-time monitoring parameter set and the update threshold of each necessary monitoring feature, or the update threshold of each necessary monitoring feature, based on the simplified transformer three-dimensional model. Further details are omitted here.

[0145] In the above embodiments, during module processing, a corresponding simplification process is generated using the model key feature dataset. Based on this dataset and the corresponding simplification process, the initial 3D model is simplified to generate a preliminary simplified 3D transformer model. This model is then simulated, and based on the simulation data, iterative optimization is selectively performed to obtain a simplified 3D transformer model. This ensures that the simplified model accurately reflects the transformer data while also optimizing the simplified model. Furthermore, using the transformer 3D model and the model key feature dataset, a corresponding real-time monitoring parameter set is generated to constitute the model simplification scheme. This achieves automatic generation of the model simplification scheme, ensuring that the simplification of the transformer 3D model is achieved while maintaining real-time monitoring of the transformer. It also ensures that the real-time monitoring parameter set matches the transformer 3D model, improving the intelligent simplification capability of the model simplification module for 3D models.

[0146] Step S104: Based on the real-time monitoring parameter set corresponding to the three-dimensional model of the transformer in the model simplification scheme, obtain the real-time monitoring data of the transformer;

[0147] Step S105: Substitute the real-time monitoring data of the transformer into the model simplification module, selectively adjust the model simplification scheme to achieve lightweighting of the transformer 3D model, i.e., the transformer geometric-physical digital twin model, thereby ensuring the optimization of the lightweight transformer 3D model while guaranteeing the accuracy of the transformer 3D model in reflecting real-time data.

[0148] Specifically, the step of substituting the real-time monitoring data of the transformer into the model simplification module and selectively adjusting the model simplification scheme to simplify the three-dimensional model of the transformer includes:

[0149] Based on the real-time monitoring data of the transformer, the simplified three-dimensional model of the transformer is updated to obtain a simplified three-dimensional model of the transformer mapped with the real-time monitoring data of the transformer and the processing time.

[0150] Based on the real-time monitoring data of the transformer, the initial three-dimensional model is updated to obtain an initial three-dimensional model mapped with the real-time monitoring data of the transformer and the processing time.

[0151] If the processing time of the simplified 3D transformer model is longer than that of the initial 3D model, then the model simplification scheme will be adjusted to achieve the simplification of the 3D transformer model.

[0152] Otherwise, the simplified 3D model of the transformer mapped with the real-time monitoring data of the transformer is compared with the initial 3D model of the simplified 3D model of the transformer mapped with the real-time monitoring data of the transformer to obtain the accuracy and error value of the simplified 3D model of the transformer.

[0153] If the accuracy of the simplified 3D transformer model reaches the preset accuracy threshold, and the error value of the simplified 3D transformer model is lower than the preset error threshold, then no adjustment to the model simplification scheme will be performed.

[0154] Otherwise, adjustments will be made to the model simplification scheme to simplify the three-dimensional model of the transformer.

[0155] Specifically, the preset accuracy threshold can be 90% or 95%, and the preset error threshold can be 5% or 10%. The preset accuracy threshold and preset error threshold settings here are only illustrative examples. In actual testing, those skilled in the art can set them according to actual needs, which will not be elaborated here.

[0156] Specifically, the simplified three-dimensional model of the transformer is updated using the real-time monitoring data of the transformer to obtain a simplified three-dimensional model of the transformer mapped with the real-time monitoring data. The mapping method can be a mapping method that represents parameters such as temperature, vibration, and load in the model in the form of color, shape changes, etc. Furthermore, the mapping method can adopt a low-data-volume dynamic update method. When incrementally updating, only the changed data is updated, thereby reducing the data processing overhead. The choice of mapping method here is only an example. In actual testing, those skilled in the art can choose according to actual needs, as long as it can realize the updating of the simplified three-dimensional model of the transformer based on the real-time monitoring data of the transformer to obtain a simplified three-dimensional model of the transformer mapped with the real-time monitoring data of the transformer. Further details are omitted here.

[0157] In the above embodiments, by updating the simplified 3D model of the transformer and the initial 3D model with the acquired real-time monitoring data of the transformer, the updated data is mapped to the simplified 3D model of the transformer. Furthermore, by comparing the simplified 3D model of the transformer mapped with the real-time monitoring data with the initial 3D model mapped with the real-time monitoring data of the transformer, the accuracy and real-time performance of the simplified 3D model of the transformer are verified. This achieves automatic verification of the effect of the simplified 3D model of the transformer in terms of accuracy and real-time performance, thereby ensuring that the simplified 3D model of the transformer can accurately reflect the key state of the transformer. In addition, it enables continuous adjustment and optimization of the lightweight processing parameters according to actual usage, achieving a balance between lightweight and accuracy.

[0158] According to the embodiments of this disclosure, the following technical effects are achieved:

[0159] By obtaining the model feedback requirements, a key feature dataset of the model is generated, providing the data foundation required for the model simplification module. By substituting the initial 3D model and the key feature dataset into the model simplification module, a model simplification scheme is obtained, achieving automatic generation of the simplification scheme. This ensures that the model simplification scheme simplifies the transformer's 3D model while maintaining real-time monitoring of the transformer and ensuring that the real-time monitoring parameter set matches the transformer's 3D model. Then, based on the real-time monitoring parameter set corresponding to the transformer's 3D model in the model simplification scheme, real-time monitoring data of the transformer is obtained. This real-time monitoring data is then substituted into the model simplification module, selectively adjusting the model simplification scheme to simplify the transformer's 3D model. This ensures that the simplified transformer 3D model is optimized while maintaining the accuracy of the transformer's 3D model in reflecting real-time data.

[0160] Example 2:

[0161] According to Embodiment 2 of the present invention, a lightweight method for transformer digital twin models is proposed. By utilizing the Vertex Saliency-oriented Classified Edge Collapse (VS-CEC) algorithm, a physical error sensitivity term is creatively introduced into the saliency evaluation system. This proposes a geometry-physics jointly driven lightweight method for digital twin models. While ensuring the structural integrity of the transformer digital twin model and the accuracy of key physical properties, the method reduces the amount of model data and improves visualization efficiency and simulation calculation efficiency.

[0162] Figure 2 The main process diagram of this embodiment two is shown; Figure 3 A detailed flowchart of this second embodiment is shown. Figure 2 and Figure 3 As shown, a model simplification scheme or lightweight strategy is constructed based on the model's key feature dataset and the initial 3D model; according to the lightweight strategy, the lightweighting of the transformer's geometric-physical digital twin model is achieved, including:

[0163] (1) Extraction of key information from geometric model and region labeling

[0164] Feature data is obtained by extracting features and labeling regions from the key geometric information in the model's key feature dataset.

[0165] Specifically, vertices v, edges e, and faces f, as well as mesh structure, normal vectors, and edge weights, are extracted from the 3D CAD / BIM / FEM model of the power transformer. Key structural areas such as the transformer core, windings, body, tank, and bushings are labeled as prior inputs.

[0166] (2) Vertex saliency calculation driven by geometry-physics

[0167] The vertex saliency of the geometric model is calculated using a vertex saliency-driven classification edge contraction algorithm.

[0168] In the VS-CEC method, the vertex saliency of the geometric model needs to comprehensively consider the strength of geometric features, topological importance, and structural functional relevance. This invention creatively introduces a physical error sensitivity term into the vertex saliency evaluation, extending the geometry-dominated saliency function into a geometry-physics jointly driven saliency function.

[0169]

[0170] Where S(v) represents vertex saliency, H(v) represents mean curvature, reflecting geometric abrupt changes; B(v)∈{0,1} indicates whether it is a boundary / corner point; ε is the gradient of the physical field (electromagnetic field, temperature field, etc.); Ep(v) is the local physical error sensitivity index; α, β, γ, and δ are adjustable weighting factors.

[0171] During the simplification process, the potential error of simplification at each vertex to the physics calculation is evaluated. The local physics error based on simulation bias is determined by the following formula:

[0172]

[0173] in, To simplify the values ​​of physical quantities (temperature, electric field, magnetic field, etc.) obtained by interpolation in the model, P(v) is the value of the physical quantity obtained by simulation calculation in the original model, and Pmax is the maximum reference value used for normalization.

[0174] (3) Classify all edges according to the salience of their adjacent vertices:

[0175] All edges in the feature data are classified into three categories: Category I, Category II, and Category III, based on the salience of their adjacent vertices.

[0176] Based on the vertex saliency calculation results in step (2), the edges are divided into three categories:

[0177] ① For the first type, both vertices have low significance; for this type of vertex, we should preferentially shrink it.

[0178] ② Type II, where one end is highly significant and the other end is low significant, requires restriction and contraction at the vertex.

[0179] ③ Type III: Both vertices are highly significant. For this type of vertex, contraction is refused, meaning the edge is a protected edge.

[0180] In addition to vertex saliency, edges also have a saliency index (edge ​​saliency), which is the maximum of the two vertex saliency factors.

[0181] S e (e ij )=max{S(v i ), S(v j )}

[0182] Wherein, S(v i ), S(v j ) are the significance indices of the two vertices.

[0183] (4) Construct an edge contraction cost function with significant penalty:

[0184] Constructing a geometry-physics jointly driven edge contraction cost function:

[0185]

[0186] The first term on the right-hand side of the equation is the geometric error term after shrinking the edge, where v k =|x, y, z, 1| T Let be the homogeneous coordinates of the vertices, and the traditional geometric error matrix. p k =|a k b k c k d k | T From the triangular plane a containing that vertex k x+b k y+c k z+d k =0.

[0187] The second term on the right side of the equation is the regularity penalty term. R(eij) is the regularity penalty function for the newly generated triangle after edge contraction, and λ is the adjustment weight coefficient, which is generally between 0.01 and 1.

[0188] The third term on the right side of the equation is the significance penalty term, where Se(eij) is the edge significance and μ is the significance penalty coefficient, which controls the degree of preservation of highly significant regions.

[0189] (5) Edge shrinkage based on priority queue and dynamic classification: I-type edges are shrunk preferentially according to the edge shrinkage cost function through the min-heap priority queue, and the topology and saliency data are updated synchronously until the number of faces ≤ target number or shrinkage cost ≥ error upper limit is satisfied.

[0190] Specifically: edge contraction refers to shrinking an edge e ij =(v i ,v j The two vertices v i and v j They are merged into a new vertex v', and the affected faces are deleted to form a new topology.

[0191] Construct a min-heap priority queue, and shrink the queue by taking the edge with the minimum cost each time, according to the following rules:

[0192] Rule 1: Prioritize shrinking edges of type I;

[0193] Rule 2: If the costs are equal, prioritize shrinking the edge with lower significance.

[0194] Select the edge with minimum cost from the heap, perform a shrinkage operation and update the local topology, and simultaneously update the edge classification and vertex saliency dynamically. The shrinkage process terminates when any termination condition is triggered.

[0195] Termination condition 1: Current quantity ≤ target quantity;

[0196] Termination condition 2: Edge contraction cost ≥ upper limit of error tolerance.

[0197] (6) Lightweight model post-processing and evaluation:

[0198] The lightweight model is mesh smoothed and the cost function is reconstructed if the Hausdorff distance check is used to ensure that HD(A,B)≤3%.

[0199] Specifically, the lightweight model undergoes mesh smoothing (to avoid flattening and folding) and is compared with the original model using Hausdorff Distance (HD) verification. This assesses the degree of geometric deviation between the lightweight model and the original model, ensuring that the lightweighting process does not cause serious errors in critical structures. The calculation formula can be expressed as:

[0200]

[0201] Where A = {a1, a2, ..., am} is the original model surface point set, B = {b1, b2, ..., bn} is the lightweight model surface point set, and ||ab|| represents the Euclidean distance between two points.

[0202] To ensure that the loss of details in the lightweight model is within an acceptable range, HD(A, B) ≤ 3% is specified. If this condition is not met, the edge shrinkage cost function is reconstructed.

[0203] (7) Complete the lightweighting of the three-dimensional model of the transformer (i.e., the geometric-physical digital twin model of the transformer).

[0204] According to the technical solution of this second embodiment of the present invention, it is possible to reduce the amount of model data and improve visualization efficiency and simulation calculation efficiency while ensuring the structural integrity of the digital twin model of the transformer and the accuracy of its key physical properties.

[0205] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, because according to this disclosure, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this disclosure.

[0206] Furthermore, this disclosure also provides a lightweight system for digital twin 3D models of transformers.

[0207] See appendix Figure 4 , Figure 4A block diagram of a lightweight system for a transformer digital twin 3D model is shown according to an embodiment of the present disclosure. Figure 4 As shown, the lightweight system 200 in this embodiment includes at least a communication device 203, a data acquisition device 204, a model simplification module 202, and a control device 201, respectively installed inside and outside the battery swapping cabinet. The communication device 203 is used to realize information interaction between the data acquisition device 204 and the transformer and management terminal. The data acquisition device 204 is used to acquire at least real-time monitoring data of the transformer. The model simplification module 202 is used to generate a model simplification scheme based on the model key feature dataset and the initial three-dimensional model, and selectively adjust the model simplification scheme according to the real-time monitoring data of the transformer to simplify the three-dimensional model of the transformer. The control device 201 includes a processor 2011 and a memory 2012. The memory 2012 can be configured to store program code 2013 for executing the lightweight method for the three-dimensional model of the transformer digital twin in the above method embodiment. The processor 2011 can be configured to execute the program code 2013 in the memory 2012. The program code 2013 includes, but is not limited to, the program code 2013 for executing the lightweight method for the three-dimensional model of the transformer digital twin in the above method embodiment. For ease of explanation, only the parts relevant to the embodiments of this disclosure are shown. For specific technical details not disclosed, please refer to the method section of the embodiments of this disclosure. The control device 201 may be a control device comprising various electronic devices.

[0208] Specifically, the communication device 203 can be connected via Wi-Fi, Bluetooth, or wired connection. The choice of communication device 203 is merely illustrative. Those skilled in the art can choose according to actual usage needs, as long as the communication device 203 enables mutual communication between the acquisition device 204 and the transformer and management terminal, thereby realizing information interaction between the acquisition device 204 and the transformer and management terminal. Further details are omitted here.

[0209] Specifically, the acquisition device 204 can be a variety of sensors installed around the transformer or a variety of sensors installed inside the transformer. The type of acquisition device 204 is not limited, and those skilled in the art can choose according to actual usage needs, as long as the acquisition device 204 can achieve the acquisition of real-time monitoring data of the transformer. Further details will not be elaborated here.

[0210] In one implementation, the specific function can be described in steps S101-S105.

[0211] The aforementioned lightweight system 200 is used for execution Figure 1 or Figure 3 The embodiments of the lightweight method for transformer digital twin 3D models shown are similar in technical principle, the technical problems solved, and the technical effects produced. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the lightweight system 200 can be found in the embodiments of the lightweight method for transformer digital twin 3D models, and will not be repeated here.

[0212] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the apparatus of this disclosure, the physical devices corresponding to these modules may be the processor 2011 itself, or a part of the software, hardware, or a combination of software and hardware within the processor 2011. Therefore, the number of modules shown in the figures is merely illustrative.

[0213] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0214] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A lightweight method for digital twin models of transformers, characterized in that, The method includes the following steps: The process involves: acquiring an initial 3D model; obtaining model feedback requirements and generating a key feature dataset based on these requirements; constructing a model simplification scheme or a lightweighting strategy based on the key feature dataset and the initial 3D model; wherein, constructing a model simplification scheme based on the key feature dataset and the initial 3D model includes: substituting the key feature dataset and the initial 3D model into a model simplification module to obtain a model simplification scheme; acquiring real-time monitoring data of the transformer based on the real-time monitoring parameter set corresponding to the transformer 3D model in the model simplification scheme; substituting the real-time monitoring data of the transformer into the model simplification module, selectively adjusting the model simplification scheme, or following the lightweighting strategy to achieve lightweighting of the transformer 3D model; The process of obtaining model feedback requirements and generating a dataset of key model features based on those requirements includes: Obtain model feedback requirements, wherein the model feedback requirements include at least one monitoring requirement from the user for at least one of the transformer 3D models and the priority level of each monitoring requirement; The monitoring requirement of at least one of the above is any type of parameter of the key components of the transformer casing, windings, core, oil tank, and cooling system, and the priority level of each monitoring requirement is set by the technicians themselves. Based on the model feedback requirements, determine at least one essential key feature of the model and the priority level of each essential key feature of the model. Based on at least one essential key feature of the model and the priority level of each essential key feature, non-essential key features of the model corresponding to the essential key features of the model and the priority level of the non-essential key features of the model are selectively generated. The model essential key features, the priority level of each model essential key feature, the selectively generated non-essential key features corresponding to the model essential key features, and the priority level of the model non-essential key features are filtered to obtain at least one model essential key feature and a non-essential key feature corresponding to the model essential key feature, or at least one model essential key feature after filtering. A model key feature dataset is constructed based on at least one of the selected essential key features of the model and the corresponding non-essential key features of the model, or on at least one of the selected essential key features of the model.

2. The method according to claim 1, characterized in that, The process involves filtering at least one essential model feature, the priority level of each essential model feature, selectively generated non-essential model features corresponding to the essential model features, and the priority level of the non-essential model features to obtain at least one filtered essential model feature and corresponding non-essential model features. Alternatively, the filtered essential model feature may include: When no non-essential key features corresponding to the essential key features of the model and no priority ranking of the non-essential key features of the model are generated: Based on the priority level of each essential key feature of the model, the essential key features of the model are selectively eliminated in the preliminary stage, resulting in multiple sets of essential key features of the model after preliminary elimination. Each set of essential key features of the model includes at least a number of essential key features of the model. Based on the set of essential key features of each model after initial elimination, the initial three-dimensional models are simplified and simulated to obtain simplified initial three-dimensional models. Based on each simplified initial 3D model, determine the corresponding dataset for each simplified initial 3D model; Based on the dataset corresponding to each simplified initial 3D model, the necessary key feature sets of multiple models after preliminary elimination are filtered to obtain the filtered sets of necessary key feature sets of multiple models. Based on the multiple sets of essential key features of the model after screening, select multiple essential key features from one set of essential key features of the model as at least one essential key feature of the model after screening.

3. The method according to claim 2, characterized in that, The step involves filtering at least one essential model key feature, the priority level of each essential model key feature, selectively generated non-essential model key features corresponding to the essential model key features, and the priority level of the non-essential model key features to obtain at least one filtered essential model key feature and corresponding non-essential model key features. Alternatively, the filtered at least one essential model key feature may further include: When generating non-essential key features of the model corresponding to the essential key features of the model, and the priority levels of the non-essential key features of the model: Based on the priority level of each essential key feature of the model, the essential key features of the model are selectively eliminated in the preliminary stage, resulting in multiple sets of essential key features of the model after preliminary elimination. Each set of essential key features of the model includes at least a number of essential key features of the model. Based on the priority of each model's non-essential key features and the multiple sets of model essential key features after initial elimination, at least one model non-essential key feature corresponding to the set of model essential key features is selectively generated to form a set of model key features. Each set of model key features includes at least one set of model essential key features and at least one model non-essential key feature corresponding to the set of model essential key features. Based on the key feature set of each model and the necessary key feature set of each model, the initial three-dimensional model is simplified and simulated to obtain the simplified initial three-dimensional model. Based on each simplified initial 3D model, determine the corresponding dataset for each simplified initial 3D model; Based on the dataset corresponding to each simplified initial 3D model, the necessary key feature sets of multiple models and the key feature sets of multiple models after preliminary elimination are filtered to obtain the filtered sets of necessary key feature sets of multiple models and / or sets of key feature sets of multiple models. Based on the filtered sets of multiple sets of essential key features and / or multiple sets of key features, select multiple essential key features from one set of essential key features as at least one essential key feature of the filtered model; or, select at least one essential key feature from one set of key features and the corresponding non-essential key features of the model as at least one essential key feature of the filtered model and the corresponding non-essential key features of the model.

4. The method according to claim 3, characterized in that, Based on the priority level of each essential key feature of the model, the essential key features of the model are selectively eliminated in the preliminary stage, resulting in multiple sets of essential key features of the model after preliminary elimination, including: Based on the model feedback requirements, determine the necessary priority level thresholds corresponding to the necessary key features of the model. If the priority level of the essential key features of the model is lower than the essential priority level threshold, then the essential key features of the model are screened out. Otherwise, the essential key features of the model are retained; To obtain all the essential key features of the retained model; All retained essential key features of the model are randomly matched to form multiple sets of essential key features of the model after initial elimination.

5. The method according to claim 4, characterized in that, Substituting the key feature dataset and the initial 3D model into the model simplification module, the resulting model simplification scheme includes: Based on the model's key feature dataset, a simplified process corresponding to the model's key feature dataset is generated. The simplified process includes at least multiple simplification types, the steps of each simplification type, and simplification parameters for each simplification type. Based on the model key feature dataset and the simplification process corresponding to the model key feature dataset, the initial three-dimensional model is simplified to generate a preliminary simplified three-dimensional transformer model. The simplified three-dimensional model of the transformer was simulated to obtain simulation data of the simplified three-dimensional model of the transformer. Based on the simulation data of the preliminary simplified three-dimensional transformer model, the preliminary simplified three-dimensional transformer model is selectively iteratively optimized to obtain the iteratively optimized three-dimensional transformer model, and the iteratively optimized three-dimensional transformer model is used as the simplified three-dimensional transformer model. Based on the simplified 3D model of the transformer and the dataset of key features of the model, a set of real-time monitoring parameters corresponding to the 3D model of the transformer is generated. The model simplification scheme is constructed and output; wherein the model simplification scheme includes at least a simplified three-dimensional model of the transformer and a set of real-time monitoring parameters corresponding to the three-dimensional model of the transformer.

6. The method according to claim 5, characterized in that, The simplified process for generating a corresponding dataset based on the model's key feature dataset includes: Based on whether there are non-essential key features in the model's key feature dataset that correspond to the necessary key features of the model, multiple simplification types are selectively determined; Based on multiple simplification types and key feature datasets of the model, multiple sets of primary simplification processes are generated, including: combining multiple simplification types to form multiple sets of simplification type combinations; generating multiple sets of simplification parameter combinations corresponding to each set of simplification type combinations, wherein the multiple sets of simplification parameter combinations include simplification parameters for each simplification type; and constructing multiple sets of primary simplification processes based on the multiple sets of simplification type combinations and the multiple sets of simplification parameter combinations corresponding to each set of simplification type combinations. Based on multiple sets of initial simplification processes, the initial three-dimensional model is simplified and simulated to obtain each simplified initial three-dimensional model. Based on each simplified initial 3D model, multiple sets of primary simplified processes are filtered to obtain the filtered primary simplified processes; Based on the initial 3D model corresponding to each filtered primary simplified process, determine the corresponding dataset of the initial 3D model corresponding to each filtered primary simplified process; Based on the dataset corresponding to the initial 3D model of each filtered primary simplified process, a set of filtered primary simplified processes is selected as the simplified process corresponding to the key feature dataset of the model.

7. The method according to claim 6, characterized in that, The real-time monitoring parameter set corresponding to the simplified 3D transformer model and its key feature dataset, generated from the simplified 3D transformer model, includes: Based on at least one essential key feature of the model in the model key feature dataset and the non-essential key feature of the model corresponding to the essential key feature, or at least one essential key feature of the model, determine at least one essential monitoring feature in the real-time monitoring parameter set and the non-essential monitoring feature corresponding to the essential monitoring feature, or at least one essential monitoring feature. Based on the simplified three-dimensional model of the transformer, the update thresholds for each necessary monitoring feature and each unnecessary monitoring feature in the real-time monitoring parameter set are determined, or the update threshold for each necessary monitoring feature. The real-time monitoring parameter set corresponding to the three-dimensional model of the transformer is obtained by using at least one necessary monitoring feature from the real-time monitoring parameter set and the non-necessary monitoring feature corresponding to the necessary monitoring feature, or at least one necessary monitoring feature and the update threshold corresponding to each monitoring feature.

8. The method according to claim 1, characterized in that, in, Based on the key feature dataset of the model and the initial 3D model, construct a model simplification scheme or a lightweight strategy; The lightweight strategy described above, which aims to reduce the weight of the transformer's geometric-physical digital twin model, includes: Feature data is obtained by extracting features and labeling regions from the key geometric information in the model's key feature dataset. The vertex saliency of the geometric model is calculated using a vertex saliency-driven classification edge contraction algorithm. All edges in the feature data are classified into three categories: Category I, Category II, and Category III, based on the salience of their adjacent vertices. Construct an edge contraction cost function with a significant penalty; The edges of type I are preferentially shrunk using the edge shrinkage cost function according to the min-heap priority queue, and the topology and saliency data are updated synchronously until the number of faces is less than or equal to the target number or the shrinkage cost is greater than or equal to the upper limit of error. The lightweight model is mesh smoothed, and the cost function is reconstructed if the Hausdorff distance check is used to ensure that HD(A,B)≤3%. Complete the lightweighting of the transformer's 3D model.

9. A lightweight system for a digital twin model of a transformer, characterized in that, The system includes a communication device, a data acquisition device, a model simplification module, and a control device. The communication device is used to realize information interaction between the data acquisition device and the transformer and the management terminal. The data acquisition device is used to acquire at least real-time monitoring data of the transformer. The model simplification module is used to generate a model simplification scheme based on the model's key feature dataset and the initial 3D model. Furthermore, it selectively adjusts the model simplification scheme based on the transformer's real-time monitoring data to simplify the 3D model of the transformer. The control device includes a processor and a memory. The memory is adapted to store multiple lines of program code, which are adapted to be loaded and run by the processor to execute the transformer digital twin model lightweighting method according to any one of claims 1 to 8.

Images