CN-120929251-B - Multi-scene dynamic simulation management system based on video twinning

CN120929251BCN 120929251 BCN120929251 BCN 120929251BCN-120929251-B

Abstract

The invention discloses a video twinning-based multi-scene dynamic simulation management system, in particular to the field of multi-scene simulation management, which comprises a real-time video acquisition module, a video stream intelligent analysis module, a twinning model construction module, a dynamic simulation module, an evaluation optimization module and a resource management and scheduling module. The multi-scene dynamic simulation management system based on video twinning acquires multi-level feature characterization, combines a video twinning model with first scene features to perform dynamic simulation, improves the capability of the system to reflect physical scenes according to real-time scene changes, monitors resource use conditions in real time through a resource management and scheduling module, dynamically distributes and schedules resources, reduces consumption of computing resources and storage resources when large-scale data and complex scenes are processed, generates the first scene features through processed video information, and is integrated into twinning model construction and dynamic simulation, and improves the integration depth of the video information in a simulation model.

Inventors

Tang Huzhen
LI HUI
Peng Qiusu
ZHANG XIAOSONG
CHEN XUEHUI
GAO JIAQUAN
LV YONGSHENG
HAO ZHUOJUN
DONG JIAXING
MENG QINGGUO
MA HUAN
Guo Aorigele
YANG YONG

Assignees

苏能(锡林郭勒)发电有限公司
南京科远智慧科技集团股份有限公司

Dates

Publication Date: 20260505
Application Date: 20250716

Claims (6)

1. The multi-scene dynamic simulation management system based on video twinning comprises a system operation database, a system central processing module and a user information end, and is characterized by further comprising: The real-time video acquisition module is used for responding to the video source of the video acquisition equipment group to acquire a first scene video of the video acquisition equipment; The video stream intelligent analysis module is used for performing intelligent analysis operation on the first scene video transmitted by the real-time video acquisition module, and the intelligent analysis operation is used for acquiring first scene characteristics corresponding to the first scene video; the twin model construction module is used for acquiring a video twin model based on the first scene video transmitted by the real-time video acquisition module; the dynamic simulation module is used for carrying out dynamic simulation operation on the first scene characteristics transmitted by the video stream intelligent analysis module based on the video twin model transmitted by the twin model construction module to acquire a second scene video; the evaluation optimization module is used for evaluating the second scene video transmitted by the dynamic simulation module and acquiring second scene characteristics corresponding to the second scene video: The method comprises the steps of acquiring real-time video stream data and video streams output after dynamic simulation by a video acquisition equipment group, integrating the real-time video stream data and the video streams to a standardized space-time coordinate system, establishing comprehensive evaluation indexes, performing parallel evaluation by three methods of physical rule verification, statistical process control and deep learning anomaly detection, determining an optimization objective function based on key optimization elements, and updating twin model driving parameters of a video twin model; before obtaining the second scene characteristics corresponding to the second scene video, performing parallel evaluation by three methods of physical rule verification, statistical process control and deep learning anomaly detection, wherein the method comprises the steps of comparing the operation parameters of key equipment in the second scene video after dynamic simulation by an energy conservation equation to obtain the scores of the physical rule verification, and constructing The Hotelling control diagram is used for evaluating the running state of key equipment in the second scene video after dynamic simulation, obtaining scores of statistical process control; the resource management and scheduling module is used for managing and scheduling the resource use of the dynamic simulation operation according to the second scene characteristics transmitted by the evaluation optimization module; The system operation database comprises all data texts of the multi-scene dynamic simulation management system, information texts output by all modules are collected in real time, the system central processing module is used for controlling information text instructions output by all modules in the system, and the user information end is information output equipment for receiving the multi-scene dynamic simulation management system.
2. The video twinning-based multi-scene dynamic simulation management system of claim 1, wherein the video stream intelligent analysis module is configured to obtain a first scene feature corresponding to a first scene video specifically comprises: Noise reduction processing is carried out on the first scene video, and noise interference caused by noise of electronic elements of equipment and environmental electromagnetic interference is removed; Adjusting the brightness, contrast and saturation of the first scene video through an image enhancement algorithm; Normalizing the video frame rate to a preset standard frame rate through a frame rate conversion algorithm; identifying power generation equipment corresponding to the first scene video by adopting a deep learning algorithm, and tracking the motion trail of the equipment by combining a Kalman filtering algorithm to acquire the running dynamics of the equipment in real time; performing feature extraction on the processed video frames by using a convolutional neural network to generate feature vectors of the first scene video; and classifying and labeling the feature vectors of the identified first scene video by using preset classification and labeling rules to generate first scene features.
3. The video twinning-based multi-scene dynamic simulation management system according to claim 2, wherein the video stream intelligent analysis module generates a feature vector of a first scene video before acquiring the first scene feature corresponding to the first scene video, and specifically comprises: Resolution based on video frames is A kind of electronic device The image, tensor, is noted as , Expressed as real space, i.e. the value of each element in the tensor belongs to real space, calculate the th Layer convolution operation output Characteristic diagram The method is specifically expressed as follows: , Wherein, the Represented as LeakyReLU activation functions, Denoted as the first A set of layer feature maps is provided, Represented as a collection The index of the feature map of the middle, Denoted as the first Layer 1 The characteristic map of the image is shown in the figure, Represented as a convolution operation, Denoted as the first Layer slave Selected from the layers Input feature map to the first The convolution kernels of the feature maps are output, Denoted as the first Layer number Bias terms for the feature map; Compression of spatial dimensions into a global average pooling Generating a feature vector after global average pooling The method is specifically expressed as follows: , Wherein, the Represented as in-channel The feature vector of the above-mentioned model, Expressed as the number of channels and the number of channels, Denoted as the first The height of the layer profile, Denoted as the first The width of the layer profile, Denoted as the first Index of the height of the layer feature map, Denoted as the first Index of the width of the layer profile, Denoted as the first Layer number The feature map is located at A value of (a); mapping the feature vector after global average pooling into a low-dimensional vector through a full connection layer, and reducing the feature vector after dimension reduction The method is specifically expressed as follows: , Wherein, the Weight matrix expressed as full connection layer for connecting Mapping to The space is provided with a plurality of grooves, Represented as a global average pooled feature vector, Expressed as bias vector of full connection layer, feature vector after dimension reduction , Expressed as a post-dimensionality reduction feature vector Is a dimension of (c).
4. The video twinning-based multi-scene dynamic simulation management system of claim 1, wherein the twinning model construction module obtains a video twinning model, and the system specifically comprises: Performing multi-mode feature decoupling operation on the first scene video to obtain video stream features corresponding to the first scene video, wherein the video stream features corresponding to the first scene video are multi-level feature characterization corresponding to the first scene video, and the multi-level feature characterization comprises space-time features, semantic features and physical attribute features; Acquiring a fusion strategy corresponding to the multi-level feature characterization; obtaining twin model driving parameters corresponding to the multi-level feature characterization according to the fusion strategy; and constructing the video twin model according to the twin model driving parameters.
5. The video twinning-based multi-scene dynamic simulation management system of claim 1, wherein the evaluation optimization module comprises a tracking error term, a control smoothing term and a complexity regularization term.
6. The video twinning-based multi-scenario dynamic simulation management system of claim 1, wherein the resource management and scheduling module comprises a resource monitoring unit, a resource demand analysis unit, a resource allocation and scheduling unit and a resource release and recovery unit; The resource monitoring unit is used for monitoring the use condition of the computing resource and the storage resource in the dynamic simulation process in real time and acquiring the use data of the resource; the resource demand analysis unit is used for evaluating the priority of the simulation calculation task according to the data quantity threshold and the simulation calculation complexity threshold; the resource allocation and scheduling unit is used for dynamically allocating and scheduling resources according to the priority of the simulation calculation task and the resource allocation proportion parameter; the resource releasing and recovering unit is used for releasing and recovering the resources in time after the simulation calculation task is completed.

Description

Multi-scene dynamic simulation management system based on video twinning Technical Field The invention relates to the technical field of multi-scene simulation management, in particular to a multi-scene dynamic simulation management system based on video twinning. Background Along with the rapid development of information technology, the Internet of things and industrial automation, efficient and accurate management and dynamic simulation of various scenes become particularly important in various fields. By simulating and predicting dynamic changes in different scenes, real restoration and dynamic management of the scenes are realized, resource allocation is optimized, and operation efficiency and safety are improved. The existing simulation management system is generally based on the traditional mathematical modeling and computer simulation technology, and dynamic simulation of different scenes is realized by collecting static data of physical entities, constructing scene models, setting parameters, running simulation and the like in a virtual space. However, when the system is actually used, the system still has some defects, such as low resource utilization rate, and low processing efficiency, which often needs to consume a large amount of computing resources and storage resources when processing large-scale data and complex scenes, the system is usually based on a static three-dimensional model, has low data updating frequency and limited coverage range, so that the system can not reflect physical scenes according to the change of real-time scenes, and has larger deviation between simulation models and actual conditions, and in the aspect of video processing, only video monitoring is performed, and video information is not deeply integrated into the simulation models. The multi-scene dynamic simulation management system based on video twinning effectively solves the problems by introducing real-time video data and digital twinning, improves the accuracy and instantaneity of a simulation model, improves the resource utilization rate, and enhances the interactivity and self-adaption capability of the system. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides a video twinning-based multi-scene dynamic simulation management system, which solves the problems in the background art through the following scheme. In order to achieve the above purpose, the present invention provides the following technical solutions: the multi-scene dynamic simulation management system based on video twinning comprises a system operation database, a system central processing module and a user information end, and further comprises: The real-time video acquisition module is used for responding to the video source of the video acquisition equipment group to acquire a first scene video of the video acquisition equipment; The video stream intelligent analysis module is used for performing intelligent analysis operation on the first scene video transmitted by the real-time video acquisition module, and the intelligent analysis operation is used for acquiring first scene characteristics corresponding to the first scene video; the twin model construction module is used for acquiring a video twin model based on the first scene video transmitted by the real-time video acquisition module; the dynamic simulation module is used for carrying out dynamic simulation operation on the first scene characteristics transmitted by the video stream intelligent analysis module based on the video twin model transmitted by the twin model construction module to acquire a second scene video; The evaluation optimization module is used for evaluating the second scene video transmitted by the dynamic simulation module and acquiring second scene characteristics corresponding to the second scene video; the resource management and scheduling module is used for managing and scheduling the resource use of the dynamic simulation operation according to the second scene characteristics transmitted by the evaluation optimization module; The system operation database comprises all data texts of the multi-scene dynamic simulation management system, information texts output by all modules are collected in real time, the system central processing module is used for controlling information text instructions output by all modules in the system, and the user information end is information output equipment for receiving the multi-scene dynamic simulation management system. Preferably, the intelligent video stream analyzing module obtains a first scene feature corresponding to the first scene video, which specifically includes: Noise reduction processing is carried out on the first scene video, and noise interference caused by noise of electronic elements of equipment and environmental electromagnetic interference is removed; Adjusting the brightness, contrast and saturation of the first scene video through an image enhancement al