An energy equipment health state intelligent evaluation system based on multi-source data fusion
Patent Information
- Application Number
- CN202610906321.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-23
- Publication Date
- 2026-09-11
AI Technical Summary
[0004]本发明的目的在于提供一种基于多源数据融合的能源设备健康状态智能评估系统，解决现有技术中能源设备监测感知维度单一、异构数据信息壁垒、评估模型适配性差、缺乏故障预测能力且系统无闭环迭代机制的问题
Smart Images
Figure CN122734646A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of energy equipment monitoring and artificial intelligence technology, specifically to an intelligent assessment system for the health status of energy equipment based on multi-source data fusion. Background Technology
[0002] As a core pillar of the national economy, the energy industry, including substations and power production workshops, operates under high loads and multiple conditions. The health of its equipment directly determines the safety, stability, and efficiency of energy production. Current energy equipment health monitoring primarily employs single-sensor detection, collecting only partial operational data. This results in missing sensor dimensions and weak data correlation, making it difficult to comprehensively reflect the true operating status of the equipment. Furthermore, existing assessment algorithms are mostly general-purpose models, lacking customization and optimization for the specific operating conditions of the energy sector, such as load fluctuations, equipment aging, and complex and variable environments. These models suffer from poor adaptability, low inference efficiency, and are prone to misdiagnosis and missed faults. Moreover, assessment methods often remain at the post-event diagnostic level, lacking the ability to predict potential faults.
[0003] Furthermore, the heterogeneity of various data types at energy production sites, including sensor data, visual images, and maintenance texts, creates information silos between these independent data systems, hindering in-depth cross-dimensional data fusion and analysis. This results in single-source and one-sided indicators for equipment health assessment. Simultaneously, existing monitoring systems lack closed-loop management of data across different stages, preventing assessment results from guiding model optimization. The systems suffer from insufficient self-iteration and self-optimization capabilities, making it difficult to adapt to the health monitoring needs of energy equipment throughout its entire lifecycle. Therefore, this articulates proposes an intelligent health status assessment system for energy equipment based on multi-source data fusion to address the problems mentioned in the background. Summary of the Invention
[0004] The purpose of this invention is to provide an intelligent health status assessment system for energy equipment based on multi-source data fusion, which solves the problems of single sensing dimension of energy equipment monitoring, heterogeneous data information barriers, poor adaptability of assessment models, lack of fault prediction capability and lack of closed-loop iterative mechanism in the existing technology.
[0005] This invention provides the following technical solution: an intelligent health status assessment system for energy equipment based on multi-source data fusion, comprising a data perception layer, a model training layer, a data fusion analysis layer, and a health assessment layer that sequentially realize bidirectional data interaction. Each layer collaborates to achieve high-precision positioning of the energy production site, full-dimensional perception of equipment status, customized model training, multi-source data fusion analysis, and intelligent assessment of equipment health status. The specific implementation steps are as follows:
[0006] S1: Through the data perception layer, high-precision positioning of the energy production site and full-dimensional data collection of the operating status of energy equipment are completed, and standardized multi-dimensional perception data are output after preprocessing.
[0007] S2: Through the model training layer, based on the AI large model architecture, the training and iterative optimization of customized operation and maintenance models for the energy industry are completed, and operation and maintenance models adapted to complex scenarios in the energy field are output.
[0008] S3: Receives standardized multi-dimensional sensing data and multi-source heterogeneous operation and maintenance data from energy sites through the data fusion and analysis layer, and outputs cross-dimensional correlation analysis results after fusion and analysis;
[0009] S4: Based on the fusion analysis results and customized operation and maintenance model, the health assessment layer completes the collaborative assessment of the health status of energy equipment through multiple indicators and outputs the equipment health status assessment results.
[0010] As a preferred embodiment of the above technical solution, step S1 specifically comprises:
[0011] S101: The data perception layer integrates the lidar acquisition module, the depth vision acquisition module, and the inertial navigation data acquisition module, and integrates a four-dimensional sensing system consisting of a visual sensing unit, an ultraviolet sensing unit, an ultrasonic sensing unit, and a gas sensing unit.
[0012] S102: Achieve high-precision positioning of energy production sites through LiDAR, depth vision and inertial navigation data acquisition modules; collect all-dimensional status data of energy equipment appearance defects, electrical discharge, mechanical noise and gas leakage through a four-dimensional sensing system, and integrate them to form multi-dimensional raw perception data.
[0013] S103: Perform noise reduction, normalization, and temporal-spatial synchronization calibration preprocessing on the multi-dimensional raw sensing data in sequence, output standardized multi-dimensional sensing data and transmit it to the data fusion analysis layer.
[0014] As a preferred embodiment of the above technical solution, step S2 specifically comprises:
[0015] S201: The model training layer builds an AI intelligent algorithm framework and constructs an energy operation and maintenance sample library that includes energy field operating condition data, equipment full-dimensional status data, equipment failure case data, and load fluctuation and equipment aging correlation data.
[0016] S202: Based on the AI intelligent algorithm framework, it calls the energy operation and maintenance sample library data, conducts special training for complex scenarios such as load fluctuations and equipment aging in the energy field, and generates an initial customized operation and maintenance model for the energy industry.
[0017] S203: Based on the equipment health status assessment results fed back by the health assessment layer and the actual operation and maintenance data at the energy site, iteratively optimize the initial customized operation and maintenance model for the energy industry, and output an adapted customized operation and maintenance model for the energy industry.
[0018] As a preferred embodiment of the above technical solution, step S3 specifically comprises:
[0019] S301: The data fusion analysis layer is equipped with a multi-source data integration and analysis algorithm developed based on the deep reasoning capability of large models, and is configured with a standardized access interface for multi-source data;
[0020] S302: Receives standardized multi-dimensional sensing data output from the data perception layer, as well as heterogeneous operation and maintenance data such as visual, audio, and text from the energy production site, through the multi-source data standardization access interface.
[0021] S303: Through multi-source data integration and analysis algorithms, standardized multi-dimensional perception data and heterogeneous operation and maintenance data are cross-dimensionally correlated and fused, breaking down information barriers between heterogeneous data, outputting multi-source data fusion analysis results, and synchronizing them to the model training layer and health assessment layer.
[0022] As a preferred embodiment of the above technical solution, step S4 specifically comprises:
[0023] S401: The health assessment layer constructs an equipment health assessment index system adapted to the customized operation and maintenance model of the energy industry. The assessment index system adjusts the assessment index and its corresponding weight in real time based on the adaptive optimization results of the model's dynamic operating conditions.
[0024] S402: Receives the multi-source data fusion analysis results output by the data fusion analysis layer, calls the energy industry customized operation and maintenance model optimized by the model training layer, and performs multi-indicator collaborative calculation and analysis of the operating status of energy equipment;
[0025] S403: Based on the results of multi-indicator collaborative analysis, the equipment health status assessment results are generated in the form of a combination of quantitative values, health level indicators, and fault warning prompts. At the same time, the assessment results are fed back to the model training layer for model iteration and synchronized to the energy field operation and maintenance terminal.
[0026] As a preferred embodiment of the above technical solution, the large model is the DeepSeek large model, and the multi-source data integration and analysis algorithm integrates the equipment fault prediction sub-algorithm. In step S3, based on the multi-source data fusion analysis results and the customized operation and maintenance model of the energy industry, the fault prediction sub-algorithm calculates the probability of energy equipment fault occurrence, identifies potential fault types, and synchronizes the fault prediction results to the health assessment layer and incorporates them into the equipment health status assessment results.
[0027] As a preferred embodiment of the above technical solution, the positioning error range of the lidar, depth vision and inertial navigation data acquisition module is adapted to the equipment layout accuracy requirements of the energy production site's booster station and power production workshop. The detection parameters of the four-dimensional sensing system cover the operational status characteristics of the energy equipment throughout its entire life cycle, and the temporal and spatial resolutions of the acquired data match the needs of real-time monitoring of the energy equipment.
[0028] As a preferred embodiment of the above technical solution, a data storage layer is included. The data storage layer achieves bidirectional data interaction with the data perception layer, model training layer, data fusion analysis layer, and health assessment layer. It is used to store the original data, preprocessed data, intermediate calculation data, trained model files, and equipment health status assessment results of each layer, providing full data support for the root cause analysis of energy equipment failures, historical data tracing, and continuous model iteration.
[0029] As a preferred embodiment of the above technical solution, the AI intelligent algorithm framework is a customized algorithm framework for the energy field, which is adapted to the dynamic operating conditions of energy production sites, supports lightweight deployment and rapid inference calculation of models, and meets the real-time and accuracy requirements of energy equipment health status assessment.
[0030] Compared with the prior art, the beneficial effects of the present invention are:
[0031] This invention constructs a multi-level collaborative architecture comprising a data perception layer, a model training layer, a data fusion and analysis layer, and a health assessment layer. At its core, it integrates LiDAR, depth vision, inertial navigation data, and a four-dimensional sensing system to achieve high-precision positioning at energy sites and collect comprehensive data on equipment appearance, electrical, mechanical, and gas conditions. This overcomes the limitations of traditional single-sensor perception, providing high-value raw data support for health assessment. Simultaneously, it builds customized operation and maintenance models for the energy industry based on a large AI model, conducting specialized training for complex scenarios such as load fluctuations and equipment aging. These models are continuously iterated and optimized based on assessment results and actual operation and maintenance data, addressing the issue of poor adaptability of general algorithms and providing precise algorithmic support for intelligent equipment diagnosis and dynamic operating condition optimization.
[0032] Based on a multi-source data integration and analysis algorithm with large-scale model deep reasoning capabilities, this algorithm breaks down information barriers between heterogeneous data such as sensor monitoring, vision, audio, and text, enabling cross-dimensional correlation and fusion analysis. This drives the upgrade of assessment models from single indicators to multi-indicator collaborative analysis, providing a more comprehensive and accurate reflection of equipment health status. Furthermore, by constructing a four-layer bidirectional linkage architecture, a closed loop of data acquisition, preprocessing, fusion analysis, health assessment, and model iteration is formed, improving analysis accuracy and ensuring long-term system operation. The integrated fault prediction sub-algorithm achieves dual functions of equipment health status assessment and fault probability prediction, as well as potential type identification, promoting the upgrade of operation and maintenance models towards "pre-event warning, in-event monitoring, and post-event analysis," and facilitating the digital transformation of energy production sites. Attached Figure Description
[0033] Figure 1 This is a flowchart illustrating the steps of an intelligent health status assessment system for energy equipment based on multi-source data fusion. Detailed Implementation
[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
[0035] Please see Figure 1 As shown, this invention provides a technical solution: an intelligent health status assessment system for energy equipment based on multi-source data fusion, comprising a data perception layer, a model training layer, a data fusion analysis layer, and a health assessment layer that sequentially realize bidirectional data interaction. Each layer collaborates to complete high-precision positioning of the energy production site, full-dimensional perception of equipment status, customized model training, multi-source data fusion analysis, and intelligent assessment of equipment health status. The specific implementation steps are as follows:
[0036] S1: Through the data perception layer, high-precision positioning of the energy production site and full-dimensional data collection of the operating status of energy equipment are completed, and standardized multi-dimensional perception data are output after preprocessing.
[0037] S2: Through the model training layer, based on the AI large model architecture, the training and iterative optimization of customized operation and maintenance models for the energy industry are completed, and operation and maintenance models adapted to complex scenarios in the energy field are output.
[0038] S3: Receives standardized multi-dimensional sensing data and multi-source heterogeneous operation and maintenance data from energy sites through the data fusion and analysis layer, and outputs cross-dimensional correlation analysis results after fusion and analysis;
[0039] S4: Based on the fusion analysis results and customized operation and maintenance model, the health assessment layer completes the collaborative assessment of the health status of energy equipment through multiple indicators and outputs the equipment health status assessment results.
[0040] Step S1 is as follows:
[0041] S101: The data perception layer integrates the lidar acquisition module, the depth vision acquisition module, and the inertial navigation data acquisition module, and integrates the visual sensing unit, the ultraviolet sensing unit, the ultrasonic sensing unit, and the gas sensing unit to form a four-dimensional sensing system.
[0042] The data perception layer integrates a lidar acquisition module, a depth vision acquisition module, and an inertial navigation data acquisition module, and combines a visual sensing unit, an ultraviolet sensing unit, an ultrasonic sensing unit, and a gas sensing unit to form a four-dimensional sensing system. Each acquisition module and sensing unit adopts a hardware clock synchronous triggering design, with the triggering synchronization error controlled within 1ms, ensuring the consistency of multi-source data in the time dimension. The lidar acquisition module realizes the acquisition of environmental point cloud data, the depth vision acquisition module completes the visual imaging and contour recognition of equipment, and the inertial navigation data acquisition module outputs spatial attitude and position data in real time. The four-dimensional sensing system realizes the full-dimensional perception of the operating status of energy equipment from four dimensions: equipment visual features, electrical discharge signals, mechanical vibration and ultrasound, and environmental gas concentration. The data acquired by each unit is aggregated to the data preprocessing module through an industrial bus, laying the foundation for subsequent data processing.
[0043] S102: Achieve high-precision positioning of energy production sites through LiDAR, depth vision and inertial navigation data acquisition modules; collect all-dimensional status data of energy equipment appearance defects, electrical discharge, mechanical noise and gas leakage through a four-dimensional sensing system, and integrate them to form multi-dimensional raw perception data.
[0044] High-precision positioning at energy production sites is achieved through the collaborative use of LiDAR, depth vision, and inertial navigation data acquisition modules. LiDAR utilizes spatial point cloud modeling for large-scale environmental positioning, depth vision achieves precise point calibration through equipment feature matching, and the inertial navigation module provides continuous attitude and position compensation. This synergistic interaction effectively avoids the positioning limitations of single modules in complex conditions such as obstruction and dust, and is suitable for the precision requirements of equipment layout in substations and power production workshops. Simultaneously, a four-dimensional sensing system collects comprehensive status data of energy equipment: visual sensors capture defects such as cracks and corrosion, ultraviolet sensors detect electrical discharge phenomena, ultrasonic sensors identify mechanical noise characteristics, and gas sensors monitor leaked gas concentrations. Positioning data is precisely correlated and integrated with the four types of sensor data according to timestamps, forming multi-dimensional raw perception data covering location, appearance, electrical, mechanical, and environmental aspects, providing a complete and reliable data source for subsequent health assessments.
[0045] S103: Perform noise reduction, normalization, and temporal-spatial synchronization calibration preprocessing on the multi-dimensional raw sensing data in sequence, output standardized multi-dimensional sensing data and transmit it to the data fusion analysis layer.
[0046] The multi-dimensional raw sensing data undergoes denoising, normalization, and temporal-spatial synchronization calibration preprocessing sequentially, outputting standardized multi-dimensional sensing data which is then transmitted to the data fusion and analysis layer. First, for different types of data, such as LiDAR point clouds, visual images, and sensor signals, appropriate filtering algorithms are used to remove environmental interference, equipment noise, and abrupt data jumps, restoring the true characteristics of the data. Then, multi-source data of different dimensions and magnitudes are normalized to unify the data value range and eliminate the impact of dimensional differences on subsequent analysis. Finally, time synchronization calibration is performed based on a unified clock reference, aligning the acquisition timestamps of each module and sensor unit. Simultaneously, spatial coordinate calibration is performed using positioning data, achieving precise matching of multi-source data in the temporal and spatial dimensions, ensuring data consistency and usability, and providing standardized, high-quality input data for subsequent multi-source data fusion analysis.
[0047] Step S2 is as follows:
[0048] S201: The model training layer builds an AI intelligent algorithm framework and constructs an energy operation and maintenance sample library that includes energy sector operating condition data, equipment full-dimensional status data, equipment failure case data, and load fluctuation and equipment aging correlation data.
[0049] The model training layer establishes an AI intelligent algorithm framework adapted to energy industry scenarios. This framework is compatible with the needs of large-scale model inference, deep learning training, and lightweight deployment, supporting the entire lifecycle development of energy operation and maintenance models. Simultaneously, an energy operation and maintenance sample library covering multi-dimensional data is constructed. The sample library is stored according to data type: it includes operating condition data at different times and under different loads at energy production sites; comprehensive status data of equipment throughout its lifecycle, including appearance, electrical, mechanical, and environmental aspects; fault case data formed from historical fault diagnosis and maintenance records; and correlation analysis data corresponding to load fluctuation patterns and equipment aging trends. The sample library establishes a dynamic update and quality verification mechanism, regularly supplementing it with newly added on-site operation and maintenance data and removing invalid and abnormal samples to ensure the integrity, accuracy, and timeliness of the data. This provides high-quality, full-scenario data source support for subsequent customized operation and maintenance model training.
[0050] S202: Based on the AI intelligent algorithm framework, it calls the energy operation and maintenance sample library data, conducts special training for complex scenarios such as load fluctuations and equipment aging in the energy field, and generates an initial customized operation and maintenance model for the energy industry.
[0051] Based on an AI-powered intelligent algorithm framework, this system utilizes the full dataset of an energy operations and maintenance (O&M) sample library for specialized training, targeting complex scenarios such as load fluctuations and equipment aging in the energy sector. This generates an initial, customized O&M model for the energy industry. The training process centers on real-world O&M data from energy sites, focusing on reinforcement learning for typical complex operating conditions such as drastic load fluctuations and long-term equipment aging. This allows the model to fully learn the changing patterns and fault characteristics of equipment states under different scenarios, deeply adapting to the specific application needs of the energy industry. A layered training strategy is employed: first, a basic model is pre-trained based on the full dataset; then, fine-tuning is performed for specific scenarios such as load fluctuations and equipment aging. This enhances the model's feature extraction and decision-making capabilities under complex operating conditions, ultimately generating an initial customized O&M model with intelligent equipment diagnosis and dynamic adaptive optimization capabilities, laying the algorithmic foundation for subsequent system iteration and optimization.
[0052] S203: Based on the equipment health status assessment results fed back by the health assessment layer and the actual operation and maintenance data at the energy site, iteratively optimize the initial customized operation and maintenance model for the energy industry, and output an adapted customized operation and maintenance model for the energy industry.
[0053] Based on the equipment health status assessment results from the health assessment layer and the actual operation and maintenance data from the energy site, the initial customized operation and maintenance model for the energy industry is continuously iterated and optimized. The final output is an adapted customized operation and maintenance model for the energy industry. The model diagnostic biases and unidentified fault characteristics from the assessment results, as well as newly added actual operation and maintenance data such as load fluctuations and equipment aging, are used as incremental training samples to supplement the energy operation and maintenance sample library. Online incremental training is used to fine-tune the initial model, specifically strengthening its feature recognition and decision-making capabilities in actual working conditions, correcting model assessment biases, and optimizing the model's internal parameters and inference logic based on model adaptation issues reported from on-site operation and maintenance. This ensures the model continuously meets the actual operational needs of the energy site, constantly improving the model's diagnostic accuracy and adaptability in complex scenarios.
[0054] Step S3 is as follows:
[0055] S301: The data fusion and analysis layer is equipped with a multi-source data integration and analysis algorithm developed based on the deep reasoning capability of large models, and is configured with a standardized access interface for multi-source data;
[0056] The data fusion and analysis layer is equipped with a multi-source data integration and analysis algorithm developed based on large-model deep inference capabilities, and features a standardized multi-source data access interface. This algorithm is deeply adapted to the characteristics of multi-source heterogeneous data in the energy sector, and can uncover correlations between different types of data, such as sensor monitoring, visual and audio data, and text-based maintenance data, achieving deep fusion and intelligent analysis of cross-dimensional data. The standardized access interface is compatible with mainstream industrial communication protocols such as Modbus and OPCUA, supporting seamless access from different terminals such as LiDAR, 4D sensing, and field maintenance systems. It can unify the format and convert protocols for various input data, breaking down data barriers between devices and systems, ensuring efficient and stable aggregation of multi-source data, and providing a standardized and regulated data input foundation for subsequent data fusion and analysis.
[0057] S302: Receives standardized multi-dimensional sensing data output from the data perception layer, as well as heterogeneous operation and maintenance data such as visual, audio, and text from the energy production site, through the multi-source data standardization access interface.
[0058] Through a standardized multi-source data access interface, standardized multi-dimensional sensing data output from the data perception layer is received synchronously, along with heterogeneous operation and maintenance data from energy production sites, including visual, audio, and textual data. The standardized multi-dimensional sensing data encompasses high-precision on-site positioning data and full-dimensional equipment status data acquired through four-dimensional sensing, serving as the core foundational data for equipment health assessment. The heterogeneous operation and maintenance data includes visual images from equipment inspections, audio signals from equipment operation, on-site operation and maintenance records, and textual information from fault reports. The interface, through a unified data format and communication protocol, enables efficient aggregation and unified reception of data from different types and sources. Simultaneously, it performs real-time verification of the received data, eliminating invalid and missing data to ensure the integrity and validity of the accessed data, thus preparing the data for subsequent in-depth multi-source data fusion analysis.
[0059] S303: Through multi-source data integration and analysis algorithms, standardized multi-dimensional perception data and heterogeneous operation and maintenance data are cross-dimensionally correlated and fused, breaking down information barriers between heterogeneous data, outputting multi-source data fusion analysis results, and synchronizing them to the model training layer and health assessment layer.
[0060] This algorithm integrates and analyzes standardized multi-dimensional sensing data and heterogeneous operation and maintenance data across dimensions, deeply exploring the inherent correlations between different types of data. It breaks down information barriers between heterogeneous data, accurately outputs multi-source data fusion analysis results, and simultaneously pushes them to the model training and health assessment layers. The algorithm employs a feature extraction and correlation matching fusion strategy for different data formats, including sensor monitoring, visual images, audio signals, and text records. It first transforms unstructured data into standardized feature vectors, then performs dimensional alignment and deep fusion with structured sensing data, maximizing the value of multi-source data. The fusion analysis results retain the core features of equipment status while integrating full information from the operation and maintenance process, providing comprehensive and in-depth data analysis support for accurate judgment in the health assessment layer and iterative optimization in the model training layer.
[0061] Step S4 is as follows:
[0062] S401: The health assessment layer constructs an equipment health assessment index system that is adapted to the customized operation and maintenance model of the energy industry. The assessment index system is adaptively optimized based on the dynamic operating conditions of the model, and the assessment index and corresponding weights are adjusted in real time.
[0063] The health assessment layer constructs an equipment health assessment indicator system deeply adapted to the customized operation and maintenance model of the energy industry. This system revolves around the operating characteristics of energy equipment, covering core dimensions such as core operating parameters, environmental influencing factors, potential faults, and equipment aging trends. It includes multiple sub-indicators, forming a clearly hierarchical and comprehensive assessment system. This indicator system is not fixed but dynamically optimized based on the model's dynamic operating conditions and changes in actual operating conditions at the energy site. The system dynamically filters and adjusts each indicator, while simultaneously optimizing the weight allocation of each indicator in real time. This ensures the assessment system always closely reflects the actual operating status of the equipment, guaranteeing the accuracy, adaptability, and scientific rigor of the health assessment.
[0064] S402: Receives the multi-source data fusion analysis results output by the data fusion analysis layer, calls the energy industry customized operation and maintenance model optimized by the model training layer, and performs multi-indicator collaborative calculation and analysis of the operating status of energy equipment;
[0065] The health assessment layer receives multi-source data fusion analysis results from the data fusion analysis layer in real time. Through a standardized data interaction interface, it retrieves the energy industry-customized operation and maintenance model optimized by the model training layer. Leveraging the model's intelligent analysis capabilities, it performs multi-indicator collaborative calculations and in-depth analysis of the energy equipment's operating status. The calculation and analysis process uses the fused, full-dimensional data as its core basis, combined with an adapted equipment health assessment indicator system, to accurately quantify various primary and secondary indicators. Simultaneously, it correlates historical equipment operating data with similar fault characteristics, achieving a multi-dimensional and comprehensive assessment of the equipment's current operating status. This fully uncovers the underlying patterns in the equipment's condition, providing a scientific and accurate quantitative analysis basis for subsequent health level determination.
[0066] S403: Based on the results of multi-indicator collaborative analysis, the equipment health status assessment results are generated in the form of a combination of quantitative values, health level indicators, and fault warning prompts. At the same time, the assessment results are fed back to the model training layer for model iteration and synchronized to the energy field operation and maintenance terminal.
[0067] Based on the results of multi-indicator collaborative analysis, the health assessment layer generates a complete equipment health status assessment result in the form of a combination of quantitative values, health level labels, and fault early warning prompts. The quantitative values intuitively reflect the overall health level of the equipment, the health level labels are divided into gradient levels according to the scores, and the fault early warning prompts accurately indicate potential fault points, risk levels, and scope of impact. This assessment result is fed back to the model training layer in real time, serving as incremental samples to support the continuous iterative optimization of the model, allowing the model's diagnostic capabilities to be deeply adapted to on-site operating conditions. On the other hand, it is simultaneously pushed to the energy site operation and maintenance terminal through the Industrial Internet of Things, clearly displaying the equipment health details. This provides operation and maintenance personnel with accurate and intuitive decision-making basis for fault handling, inspection planning, and maintenance arrangements, improving the efficiency and pertinence of on-site operation and maintenance work.
[0068] As one implementation method in this embodiment, the large model is the DeepSeek large model, and the multi-source data integration and analysis algorithm integrates the equipment fault prediction sub-algorithm. In step S3, based on the multi-source data fusion analysis results and the customized operation and maintenance model of the energy industry, the fault prediction sub-algorithm calculates the probability of energy equipment fault occurrence, identifies potential fault types, and synchronizes the fault prediction results to the health assessment layer and incorporates them into the equipment health status assessment results.
[0069] The DeepSeek large model is selected, relying on its powerful deep reasoning and feature mining capabilities to provide algorithmic support for multi-source data integration and analysis. The multi-source data integration and analysis algorithm integrates a sub-algorithm for equipment fault prediction. In the data analysis process of step S3, this sub-algorithm, based on the results of multi-source data fusion analysis and combined with the trained and optimized customized operation and maintenance model for the energy industry, accurately calculates the probability of occurrence of various faults in energy equipment by mining fault correlation features in equipment status data. At the same time, it matches the fault feature library to achieve intelligent identification of potential fault types and synchronizes the complete fault prediction results to the health assessment layer in real time, fully incorporating the equipment health status assessment results, making the assessment conclusions more forward-looking, and providing a reliable basis for early warning of equipment faults.
[0070] As one implementation method in this embodiment, the positioning error range of the lidar, depth vision and inertial navigation data acquisition module is adapted to the equipment layout accuracy requirements of the energy production site's booster station and power production workshop. The detection parameters of the four-dimensional sensing system cover the operating status characteristics of the energy equipment throughout its entire life cycle, and the temporal and spatial resolutions of the acquired data match the needs of real-time monitoring of the energy equipment.
[0071] The collaborative positioning error control achieved by lidar, depth vision, and inertial navigation data acquisition modules is within the appropriate range, fully meeting the equipment layout monitoring accuracy requirements of scenarios such as energy production substations and power production workshops. It can accurately pinpoint the spatial location of each piece of equipment. The detection parameters of the four-dimensional sensing system comprehensively cover the operational status characteristics of energy equipment throughout its entire lifecycle, from commissioning and normal operation to aging maintenance, with no monitoring dimensions missing. Simultaneously, the temporal and spatial resolution of the acquired data are set according to industrial-grade standards, precisely matching the needs of real-time monitoring of energy equipment. This ensures the timeliness and detailed accuracy of equipment status data acquisition, providing high-quality location and status data for subsequent analysis.
[0072] As one implementation method in this embodiment, it includes a data storage layer. The data storage layer and the data perception layer, model training layer, data fusion analysis layer and health assessment layer all achieve bidirectional data interaction. It is used to store the original data, preprocessed data, intermediate calculation data, trained model files and equipment health status assessment results of each layer, and provide full data support for the root cause analysis of energy equipment failures, historical data tracing and continuous model iteration.
[0073] The system also includes a data storage layer, which enables bidirectional data interaction with the data perception layer, model training layer, data fusion analysis layer, and health assessment layer. This layer can receive data uploaded from each module in real time and retrieve historical data for retrospective use as needed. The data storage layer uniformly stores the raw perception data generated at each layer, preprocessed standardized data, intermediate fusion calculation data, trained model files, and the final equipment health status assessment results. It employs a categorized storage and index management mechanism to ensure data security, integrity, and efficient read / write operations. By aggregating data from the entire process, it provides comprehensive and reliable full-data support for subsequent in-depth analysis of the root causes of energy equipment failures, historical operational status tracing and comparison, continuous model iteration and optimization, and operation and maintenance decisions.
[0074] As one implementation method in this embodiment, the AI intelligent algorithm framework is a customized algorithm framework for the energy field, adapted to the dynamic operating conditions of energy production sites, supporting lightweight deployment and rapid inference calculation of models, and meeting the real-time and accuracy requirements of energy equipment health status assessment.
[0075] The AI-powered intelligent algorithm framework is a customized framework specifically designed for the energy sector. It is deeply adapted to the dynamic operating conditions of energy production sites, including load fluctuations, equipment aging, and complex and ever-changing environments. It can flexibly address the operation and maintenance needs of different scenarios such as substations and power production workshops. The framework optimizes the algorithm structure and computational logic, supports lightweight model deployment, and is compatible with edge computing devices in the field, eliminating the need for large-scale cloud computing power. It also possesses rapid inference capabilities, efficiently processing multi-source heterogeneous data, significantly reducing data processing and status assessment time, accurately meeting the real-time requirements of energy equipment health status assessment. Furthermore, algorithm optimization improves assessment accuracy, effectively avoiding misjudgments and omissions, providing efficient and reliable algorithmic support for energy equipment operation and maintenance decisions.
[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it.
Claims
1. An energy equipment health state intelligent evaluation system based on multi-source data fusion, characterized in that, This includes a data perception layer, a model training layer, a data fusion and analysis layer, and a health assessment layer, which sequentially realize bidirectional data interaction. Each layer collaborates to achieve high-precision positioning of the energy production site, comprehensive perception of equipment status, customized model training, multi-source data fusion and analysis, and intelligent assessment of equipment health status. The specific implementation steps are as follows: S1: Precise perception and preprocessing, outputting standard data: The data perception layer completes high-precision positioning of the energy production site and full-dimensional data collection of the operating status of energy equipment. After preprocessing, standardized multi-dimensional perception data is output. S2: Customized model training and iterative adaptation to working conditions; through the model training layer based on the AI large model architecture, the training and iterative optimization of customized operation and maintenance models for the energy industry are completed, and operation and maintenance models adapted to complex scenarios in the energy field are output. S3: Integrate heterogeneous data and output correlation analysis results: The data fusion analysis layer receives standardized multi-dimensional sensing data and multi-source heterogeneous operation and maintenance data from energy sites, and outputs cross-dimensional correlation analysis results after fusion analysis. S4: Multi-indicator collaborative assessment, outputting health results: Based on the fusion analysis results and customized operation and maintenance model, the health assessment layer completes the multi-indicator collaborative assessment of the health status of energy equipment and outputs the equipment health status assessment results. 2.The multi-source data fusion based intelligent energy equipment health state evaluation system according to claim 1, characterized in that, Step S1 is as follows: S101: The data perception layer integrates the lidar acquisition module, the depth vision acquisition module, and the inertial navigation data acquisition module, and integrates a four-dimensional sensing system consisting of a visual sensing unit, an ultraviolet sensing unit, an ultrasonic sensing unit, and a gas sensing unit. S102: Achieve high-precision positioning of energy production sites through LiDAR, depth vision and inertial navigation data acquisition modules; collect all-dimensional status data of energy equipment appearance defects, electrical discharge, mechanical noise and gas leakage through a four-dimensional sensing system, and integrate them to form multi-dimensional raw perception data. S103: Perform noise reduction, normalization, and temporal-spatial synchronization calibration preprocessing on the multi-dimensional raw sensing data in sequence, output standardized multi-dimensional sensing data and transmit it to the data fusion analysis layer. 3.The multi-source data fusion based energy equipment health state intelligent evaluation system according to claim 1, characterized in that, Step S2 is as follows: S201: The model training layer builds an AI intelligent algorithm framework and constructs an energy operation and maintenance sample library that includes energy field operating condition data, equipment full-dimensional status data, equipment failure case data, and load fluctuation and equipment aging correlation data. S202: Based on the AI intelligent algorithm framework, it calls the energy operation and maintenance sample library data, conducts special training for complex scenarios such as load fluctuations and equipment aging in the energy field, and generates an initial customized operation and maintenance model for the energy industry. S203: Based on the equipment health status assessment results fed back by the health assessment layer and the actual operation and maintenance data at the energy site, iteratively optimize the initial customized operation and maintenance model for the energy industry, and output an adapted customized operation and maintenance model for the energy industry. 4.The multi-source data fusion based intelligent energy equipment health state evaluation system according to claim 1, characterized in that, Step S3 is as follows: S301: The data fusion analysis layer is equipped with a multi-source data integration and analysis algorithm developed based on the deep reasoning capability of large models, and is configured with a standardized access interface for multi-source data; S302: Receives standardized multi-dimensional sensing data output from the data perception layer, as well as heterogeneous operation and maintenance data such as visual, audio, and text from the energy production site, through a multi-source data standardization access interface. S303: Through multi-source data integration and analysis algorithms, standardized multi-dimensional perception data and heterogeneous operation and maintenance data are cross-dimensionally correlated and fused, breaking down information barriers between heterogeneous data, outputting multi-source data fusion analysis results, and synchronizing them to the model training layer and health assessment layer. 5.The multi-source data fusion based intelligent energy equipment health state evaluation system according to claim 1, wherein, Step S4 is as follows: S401: The health assessment layer constructs an equipment health assessment index system adapted to the customized operation and maintenance model of the energy industry. The assessment index system adjusts the assessment index and its corresponding weight in real time based on the adaptive optimization results of the model's dynamic operating conditions. S402: Receives the multi-source data fusion analysis results output by the data fusion analysis layer, calls the energy industry customized operation and maintenance model optimized by the model training layer, and performs multi-indicator collaborative calculation and analysis of the operating status of energy equipment; S403: Based on the results of multi-indicator collaborative analysis, the equipment health status assessment results are generated in the form of a combination of quantitative values, health level indicators, and fault warning prompts. At the same time, the assessment results are fed back to the model training layer for model iteration and synchronized to the energy field operation and maintenance terminal. 6.The multi-source data fusion based intelligent energy equipment health state evaluation system according to claim 4, characterized in that, The large model is the DeepSeek large model. The multi-source data integration and analysis algorithm integrates the equipment fault prediction sub-algorithm. In step S3, based on the multi-source data fusion analysis results and the customized operation and maintenance model of the energy industry, the fault prediction sub-algorithm calculates the probability of energy equipment failure, identifies potential fault types, and synchronizes the fault prediction results to the health assessment layer and incorporates them into the equipment health status assessment results.
7. The intelligent health status assessment system for energy equipment based on multi-source data fusion according to claim 2, characterized in that, The positioning error range of the lidar, depth vision and inertial navigation data acquisition module is adapted to the equipment layout accuracy requirements of energy production sites such as booster stations and power production workshops. The detection parameters of the four-dimensional sensing system cover the operating status characteristics of energy equipment throughout its entire life cycle, and the temporal and spatial resolution of the acquired data matches the needs of real-time monitoring of energy equipment.
8. The intelligent assessment system for the health status of energy equipment based on multi-source data fusion according to any one of claims 7, characterized in that, It includes a data storage layer, which enables bidirectional data interaction with the data perception layer, model training layer, data fusion analysis layer, and health assessment layer. It is used to store the raw data, preprocessed data, intermediate computation data, trained model files, and equipment health status assessment results of each layer, providing full data support for the root cause analysis of energy equipment failures, historical data tracing, and continuous model iteration.
9. The intelligent health status assessment system for energy equipment based on multi-source data fusion according to claim 3, characterized in that, The AI intelligent algorithm framework described is a customized algorithm framework for the energy field, adapted to the dynamic operating conditions of energy production sites, supporting lightweight deployment and rapid inference calculations of models, and meeting the real-time and accuracy requirements for energy equipment health status assessment.