CN-122020332-A - Multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation method and system

CN122020332ACN 122020332 ACN122020332 ACN 122020332ACN-122020332-A

Abstract

The invention belongs to the technical field of wind power plant fire disaster assessment, and discloses a multi-mode fusion wind power plant fire hazard multi-source data space-time synchronization assessment method and a multi-mode fusion wind power plant fire hazard multi-source data space-time synchronization assessment system, wherein a reference calibration module is used for building a digital twin simulation, double closed loop dynamic calibration and reliability quantification trinity mechanism; the time synchronization adopts a triple strategy of GPS time service, local clock compensation and transmission delay prediction, the LSTM network is combined to compensate the transmission delay in advance, the space calibration relies on a fan three-dimensional digital twin model, the installation deviation and vibration drift are corrected through visual identification and coordinate matching, a quality grading module constructs a three-dimensional quality model, weights are graded and distributed according to data quality, the evaluation deviation caused by data isomerism is reduced, a characteristic fusion module adopts a space-time characteristic, modal characteristic and quality weight cross fusion mechanism, the data time sequence association is captured through an overlapped time window, and the space association is excavated by combining the digital twin space topological relation.

Inventors

FAN SHIKAI
LI ZHUXIN
ZHANG RONGXUN
REN BIN

Assignees

贵州龙源新能源有限公司

Dates

Publication Date: 20260512
Application Date: 20260116

Claims (10)

1. The multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation system is characterized by comprising a data preprocessing module, a reference calibration module, a quality grading module, a characteristic fusion module, a hidden danger evaluation module, a decision early warning module and a self-adaptive optimization module; The data preprocessing module establishes a full-source access and edge preprocessing architecture, accesses multi-type wind farm data, completes format unification and preliminary quality improvement, and outputs standardized combined data comprising a data body, a device identifier, an original space-time tag and an edge preprocessing quality mark; The reference calibration module is used for realizing the space-time unification of the multi-source data based on the digital twin model and generating space-time reliability scores reflecting the reliability of the data; The quality grading module is used for establishing a three-dimensional quality grading model by combining the space-time credibility score and the preprocessing quality label, grading and screening the data and dynamically distributing weights; the feature fusion module is used for receiving the grading data and the weight distribution result, mining the time sequence relevance and the space relevance of the data, fusing the multi-type data features and converting the multi-type data features into feature vectors with uniform dimensions; The hidden danger assessment module is used for establishing a three-level assessment architecture of a station, a unit and a core component by taking fusion characteristics as input, realizing full-dimensional hidden danger assessment by combining a cross-level linkage mechanism, correcting the risk level of the unit and forming an assessment closed loop; The decision-making early warning module is used for receiving the evaluation result, establishing an interactive interface, realizing visual display of evaluation data, hierarchical early warning and operation and maintenance scheme recommendation, and synchronizing operation and maintenance feedback data to the self-adaptive optimization module; And the self-adaptive optimization module is used for training and evaluating the optimization model and the feature fusion network based on the operation and maintenance feedback data and iteratively updating parameters of each core module.
2. The multi-mode fused wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 1, wherein the full-source access architecture of the data preprocessing module is adapted to various industrial communication protocols and industrial standard protocol analyses, and the access core data comprises four types of pyrolytic nanoparticle monitoring data, robot movement sensing data, fixed sensor monitoring data and static background data; edge preprocessing is completed by locally deploying edge computing nodes in a fan cabin, and comprises image defogging and denoising, audio signal filtering and sensor abnormal value preliminary elimination; And realizing uniform format conversion of the heterogeneous data through a protocol adaptation gateway, and outputting the standardized combined data.
3. The multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 2 is characterized in that the reference calibration module establishes a digital twin simulation, double closed loop dynamic calibration and reliability quantification trinity mechanism; The space self-adaptive calibration is based on a three-dimensional digital twin model of a fan, a dynamic space coordinate library of sensing equipment is established, a fixed mark point in a cabin is identified through machine vision, the space drift caused by sensor installation deviation and vibration is corrected by combining a real-time simulation result of the digital twin model, and a dual strategy of combining machine vision identification and digital twin coordinate matching is adopted to realize the unification of the space coordinates of a mobile robot and a fixed sensor; the three-dimensional credibility assessment model is combined with calibration accuracy, environmental interference and equipment state parameters to generate space-time credibility scores.
4. The multi-modal fusion wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 3, wherein the three-dimensional quality classification model of the quality classification module specifically comprises three dimensions of space-time reliability, modal integrity and edge preprocessing quality, wherein the space-time reliability is determined by a space-time reliability score output by a reference calibration module, the modal integrity is comprehensively evaluated through three indexes of data deletion rate, noise intensity and outlier occupation ratio, and the edge preprocessing quality is determined by a mark output by an edge calculation node; And after grading, the data A directly participate in a fusion characteristic extraction process and distribute the highest fusion weight, the data B is used and distributed with medium fusion weight after interpolation complementation and noise filtering secondary optimization, the data C is marked as invalid and triggers sensor state early warning, and the weight distribution result is synchronized to a characteristic fusion module.
5. The multi-modal fusion wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 4, wherein the feature fusion module adopts a space-time feature, modal feature and quality weight cross fusion extraction mechanism; And setting a special algorithm to extract modal characteristics according to different data types, realizing single-mode space-time association reinforcement and cross-modal characteristic interaction through a space-time and modal cross-attention network, and outputting a unified dimension characteristic vector through a cross-domain characteristic fusion network after integrating quality weights.
6. The multi-modal fusion wind farm fire hazard multi-source data space-time synchronization assessment system according to claim 5, wherein in the three-level assessment architecture of the hidden danger assessment module, a special assessment index system is established for key components by core component level assessment, component hidden danger levels are output, and a component running state correction assessment result of a digital twin model is combined; the station level assessment adopts a fuzzy comprehensive evaluation method to calculate a regional fire risk index, and a regional association analysis sub-module is arranged to assess the hidden danger conduction risk of the adjacent unit; And the cross-level linkage mechanism realizes component level hidden danger trigger unit level rechecking and unit level high risk trigger station level early warning, and station level early warning information reversely guides parameter adjustment of a unit level and component level evaluation model.
7. The multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 6, wherein the interactive interface of the decision pre-warning module realizes three-dimensional visual display of data and evaluation results based on a wind farm digital twin model; Supporting space-time backtracking of data and dynamic curve display of evaluation indexes, linking an early warning mechanism with an evaluation level, pushing and prompting slight hidden danger through a platform, triggering audible and visual warning by serious hidden danger, pushing short messages, pushing APP, and locally warning by edge nodes, and carrying out multi-channel early warning; the intelligent operation and maintenance recommendation sub-module is used for recommending an optimal operation and maintenance scheme by combining historical operation and maintenance data and a digital twin simulation result, and synchronizing operation and maintenance feedback information to the self-adaptive optimization module.
8. The multi-modal fusion wind farm fire hazard multi-source data space-time synchronization evaluation system according to claim 7, wherein the self-adaptive optimization module establishes a full-closed loop self-adaptive mechanism of data feedback, reinforcement learning training, parameter updating and model iteration; setting a dual-updating mechanism of monthly total data iteration and real-time parameter fine adjustment, and deploying the updated model after digital twin simulation verification.
9. The multi-modal fusion wind farm fire hazard multi-source data space-time synchronization assessment system of claim 8, wherein the interactive interface supports real-time display of space-time backtracking query and assessment index dynamic curves of multi-source data.
10. The method for evaluating the multi-mode fusion wind farm fire hazard multi-source data in a space-time synchronization mode is based on the system of claim 9, and is characterized by comprising the following specific steps: the data access preprocessing stage comprises the steps of accessing four types of core data, adapting to an industrial communication protocol, finishing preprocessing through a cabin local edge computing node and converting the preprocessing into a standardized format; a time-space reference calibration stage, namely realizing time unification by adopting a triple time synchronization mechanism, completing space calibration based on a three-dimensional digital twin model, and synchronously generating a time-space credibility score; The quality grading and fusion stage is to finish data grading and weight distribution through a three-dimensional quality grading model, mine space-time associated features and extract multi-mode features, and fuse the space-time and mode cross attention network into a unified dimension feature vector; The hidden danger assessment early warning stage is based on a three-level assessment architecture combined with a digital twin simulation correction assessment result, hidden danger rechecking and early warning are realized through cross-level linkage, and a visual interactive interface pushing early warning information and operation and maintenance scheme is constructed; and in the self-adaptive optimization stage, operation and maintenance feedback data are collected, parameters of each link are optimized by adopting reinforcement learning and multi-agent algorithm, and a dual mechanism of monthly iteration and real-time updating is executed.

Description

Multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation method and system Technical Field The invention belongs to the technical field of wind power plant fire disaster assessment, and particularly relates to a multi-mode fusion wind power plant fire hazard multi-source data space-time synchronization assessment method and system. Background The scale of the wind power plant is continuously enlarged, the running life of the unit is continuously increased, the hidden fire hazard presents complex characteristics of multi-source induction, cross-regional diffusion and hidden evolution, and the multi-mode data fusion evaluation becomes the core technical direction of fire prevention and control. The current related art has the following technical problems: The wind power plant multi-mode data source dispersion comprises a robot inspection image, temperature and humidity data acquired by a fixed sensor, distributed partial discharge signals and the like, wherein equipment deployment positions and types are different, a time stamp is easy to generate obvious time deviation due to equipment clock drift, space coordinates are affected by fan vibration and installation deviation and are difficult to realize accurate alignment, an existing system generally adopts a static calibration mode, and is difficult to adapt to a wind power plant dynamic operation environment, so that the problem of space-time dislocation of multi-mode data is caused, and the accuracy of hidden danger correlation analysis is affected. The prior art usually has one-sided performance, only focuses on space-time synchronization, only can ensure consistent data time sequence, or simply performs modal fusion, ignores space-time correlation of data, fails to establish a linkage mechanism among space-time synchronization precision, modal feature reliability and hidden danger assessment weight, for example, when the same high-temperature hidden danger occurs at different positions of a cabin, risk grades of the hidden danger are subjected to differentiated assessment according to space-time attribute differences, but the prior system only performs unified scoring according to modal features, so that assessment results are not matched with actual risks. In addition, the prior art has three core defects that firstly a dynamic space-time calibration scheme based on a digital twin technology is lacked, so that the problem of space-time drift caused by fan vibration and environmental disturbance is difficult to solve, secondly a multi-mode fusion mode is simpler, a weighted superposition basic mode is mostly adopted, a depth fusion framework with cross correlation between space time and mode is not formed, thirdly, the self-adaption capability of an evaluation model is lacked, an optimization mechanism based on operation and maintenance feedback is lacked, and the wind power plant is difficult to adapt to dynamic scenes such as equipment aging, working condition change and the like in long-term operation of a wind power plant. Disclosure of Invention The invention aims to provide a multi-mode fusion wind farm fire hazard multi-source data space-time synchronization evaluation method and system, which are used for solving the problems in the background technology. In order to achieve the aim, the invention provides the technical scheme that the multi-mode fusion wind power plant fire hazard multi-source data space-time synchronization evaluation system comprises a data preprocessing module, a reference calibration module, a quality grading module, a feature fusion module, a hidden danger evaluation module, a decision early warning module and a self-adaptive optimization module; preferably, the data preprocessing module establishes a dual-layer architecture of full-source access combined with edge preprocessing, and the accessed core data comprises four types of pyrolytic nanoparticle monitoring data, robot movement sensing data, fixed sensor monitoring data and static background data; Adapting to communication protocols such as industrial Ethernet, loRa, 4G/5G and the like, supporting analysis of a plurality of industrial standard protocols of Modbus-RTU, MQTT, OPCUA, and carrying out real-time preprocessing on the acquired original data by locally deploying edge computing nodes in a fan cabin, wherein the preprocessing comprises image defogging and denoising, audio signal filtering and preliminary elimination of abnormal values of sensors; The unified format conversion of the heterogeneous data is realized through the protocol adaptation gateway, and the unified format conversion is output as a combined structure of a data body, a device identifier, an original space-time tag and an edge preprocessing quality mark, so that standardized initial data is provided for the reference calibration module. The reference calibration module establishes a digital twin simulation, double-closed-loop dynamic calibration and reliab