CN-121981707-A - Fault intelligent diagnosis type thermal power plant power operation and maintenance method and system

CN121981707ACN 121981707 ACN121981707 ACN 121981707ACN-121981707-A

Abstract

The invention discloses a fault intelligent diagnosis type thermal power plant electric power operation and maintenance method and system, in particular relates to the technical field of thermal power plant electric power operation and maintenance and intelligent diagnosis, the system adopts an edge-cloud-field three-layer collaborative architecture, comprises five modules including perception, edge processing, cloud core, field execution and optimization iteration, and is based on the system, and a closed-loop mechanism is formed by integrating federal learning, digital twin and other technologies, performing full-dimensional data acquisition pretreatment, constructing and updating a three-level twin model, performing three-mode integration fault diagnosis, performing differential operation and maintenance treatment and performing model self-optimization. The invention solves the problems of difficult tracing of faults of the traditional operation and maintenance cross equipment, contradiction between data safety and sharing and the like, the diagnosis accuracy is more than or equal to 98.5 percent, the early warning level faults are prejudged 3-6 hours in advance, the operation and maintenance cost is reduced by more than 30 percent, the operation and maintenance are promoted to actively prejudge, intelligently optimize the transformation, the stable operation of the unit is ensured, and the practicability is remarkable.

Inventors

WANG JIAYIN
LIN SHAOXIONG
LI JUNYI
ZHANG YIQIAN
DING JIE
WANG SHUN
XU XIAOSHAN

Assignees

华能国际电力股份有限公司上海石洞口第一电厂
节点互联(北京)科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260204

Claims (10)

1. The intelligent fault diagnosis type electric power operation and maintenance system of the thermal power plant is characterized by adopting an edge-cloud-field three-layer cooperative architecture, comprising a sensing module (1), an edge processing module (2), a cloud core module (3), a field execution module (4) and an optimization iteration module (5), wherein the modules are in redundant connection with an optical fiber network through a 5G industrial private network; The sensing module (1) consists of a multi-type sensor cluster (101), a log acquisition unit (102), an infrared thermal imaging unit (103) and a voiceprint acquisition unit (104) and is used for acquiring structured and unstructured data of the whole unit; The edge processing module (2) adopts an industrial grade edge gateway, deploys a differential preprocessing unit (201), a space-time alignment unit (202) and a federal learning encryption unit (203) and is used for data processing, space-time alignment and encryption hierarchical transmission; The cloud core module (3) is based on a server cluster and comprises a three-level digital twin model unit (301), a three-mode fusion diagnosis engine (302), a fault propagation rule base unit (303), a blockchain evidence storage unit (304) and a big data analysis unit (305) which are used for model construction and updating, fault diagnosis and prognosis, data evidence storage and analysis; the field execution module (4) consists of a mobile terminal (401), an execution mechanism (402) and redundant equipment (403) and is used for work order receiving, field treatment assistance and equipment switching; the optimization iteration module (5) is used for deploying a reinforcement learning unit (501) and an operation and maintenance configuration optimization unit (502) and is used for model iteration and operation and maintenance configuration optimization.
2. The intelligent fault diagnosis type power operation and maintenance method for the thermal power plant is characterized by realizing the pre-judgment, diagnosis, treatment and optimization of the cross-equipment faults of the whole unit based on the system of claim 1, and specifically comprising the following steps: S1, full-dimension multi-source data acquisition and heterogeneous preprocessing, namely constructing a full-unit sensing network through a sensing module (1) to acquire multi-type data, performing differential preprocessing and space-time alignment on the heterogeneous data by an edge processing module (2), and performing hierarchical transmission after desensitization by adopting a federal learning encryption protocol; S2, constructing and dynamically updating a whole-unit digital twin model, namely constructing a device-level-system-level-whole-plant-level three-level digital twin model through a cloud core module (3), dynamically calibrating parameters through an incremental learning algorithm, and embedding a fault propagation dynamics rule base; S3, three-mode fusion fault diagnosis and early pre-judgment, namely fusing multi-mode characteristics based on a graph attention network and a D-S evidence theory, and realizing fault grading pre-judgment and accurate positioning through a three-mode cooperative engine driven by a mechanism, data and knowledge of a cloud core module (3); s4, intelligent operation and maintenance closed loop treatment and execution, namely generating a differential operation and maintenance scheme according to the fault level, performing AR technology auxiliary field treatment by a linkage field execution module (4), and recording full-flow data through a blockchain evidence storage unit (304) to form a traceable file; And S5, self-optimizing the model and the operation and maintenance system, namely, based on operation and maintenance results and newly-added data, iterating the diagnosis model through a reinforcement learning unit (501) of an optimization iteration module (5), and optimizing operation and maintenance flow and resource allocation to form a diagnosis-operation and maintenance-optimization closed loop.
3. The method of claim 2, wherein in step S1, the plurality of types of data include electrical parameters including voltage, current, harmonic distortion, contact resistance, thermal parameters, mechanical parameters, environmental parameters, and unstructured data including device log, infrared thermal imaging map, and voiceprint signals; the differential preprocessing strategy is characterized in that the analog quantity signal is combined with the Kalman filtering through the self-adaptive wavelet packet decomposition to strip noise, the digital quantity signal captures abnormal time sequence through the state transition matrix, unstructured data is extracted through OCR and voiceprint features and is converted into structured vectors, and the time stamp error after time-space alignment is less than or equal to 10ms.
4. The method of claim 2, wherein in the step S2, the dynamic calibration mechanism of the three-level digital twin model is that real-time data of the edge processing module (2) and cloud historical data are driven cooperatively, calibration parameters are learned through increment every day, and the model is corrected by combining equipment maintenance records and component replacement information, so that the consistency deviation of the model and actual equipment is less than or equal to 2%.
5. The method of claim 2, wherein in step S2, the fault propagation dynamics rule base is constructed based on a fault tree and event tree two-way deducing mechanism, so that the linkage effect of a single fault on cross-equipment can be simulated, and a fault propagation path map is generated.
6. The method of claim 2, wherein in the step S3, the three-mode collaborative engine works in a manner that a mechanism driving module judges a mechanism definitely to be faulty based on a digital twin model, a data driving module adopts a lightweight transform model to cooperatively optimize through federal learning, a knowledge driving module adjusts a membership function based on an expert rule base and fuzzy logic, and the diagnosis accuracy is more than or equal to 98.5%.
7. The method of claim 2, wherein in the step S3, the fault classification pre-judging is specifically that the early warning level fault is pre-judged 3-6 hours in advance, the emergency level fault achieves millisecond level response and positioning, the positioning coordinate error is less than or equal to 5cm, and the fault influence range and the evolution trend are synchronously output.
8. The method according to claim 2, wherein in the step S4, the differentiated operation and maintenance scheme comprises the steps that the attention level triggers a regular inspection reminder, the early warning level generates a preventive maintenance work order containing a component replacement period and an operation step, the emergency level automatically triggers a redundant equipment (403) switching and stopping scheme of the on-site execution module (4), and the AR technology provides virtual operation guidance.
9. The method of claim 2, wherein in step S4, the certification content of the blockchain certification unit (304) includes fault diagnosis results, operation and maintenance worksheets, treatment processes and equipment state changes, and a non-tamperable full life cycle operation and maintenance file is formed to support compliance audit.
10. The method of claim 2, wherein in step S5, the operation and maintenance system self-optimization is specifically that the operation and maintenance work order efficiency, cost and fault recurrence rate are analyzed through big data, the inspection period is adjusted, the stock of spare parts is optimized, and the operation and maintenance cost is reduced by more than 30%.

Description

Fault intelligent diagnosis type thermal power plant power operation and maintenance method and system Technical Field The invention relates to the technical field of power operation and maintenance and intelligent diagnosis of a thermal power plant, in particular to a fault intelligent diagnosis type power operation and maintenance method and system of the thermal power plant. Background The thermal power plant is used as an energy supply core facility, the unit structure of the thermal power plant is complex, a plurality of related components such as a boiler, a steam turbine, a generator, an electric control system and the like are covered, and the operation stability directly determines the energy supply safety and the economy. With the development of intelligent technology, the operation and maintenance of a thermal power plant is changed from a traditional periodic maintenance mode to intelligent prejudgment type operation and maintenance, and the core requirements are that the accurate diagnosis, early warning and efficient treatment of cross-equipment faults are realized, and meanwhile, the data safety and the operation and maintenance cost control are considered. The traditional periodic overhaul relies on manual experience, fault prejudgement is delayed, cross-equipment cascading faults are difficult to trace, unplanned shutdown is easy to cause, operation and maintenance cost is raised, safety risks are aggravated, part of intelligent schemes only adopt single mechanism or data driving diagnosis, multi-mode data fusion capability is lacked, diagnosis precision and hidden fault identification effect are poor, cross-equipment fault propagation influence cannot be effectively simulated, in addition, multi-plant data collaborative optimization is easy to leak equipment operation privacy, the traditional scheme lacks a 'diagnosis-treatment-optimization' closed-loop mechanism, the cooperation among all modules of an operation and maintenance system is poor, data interaction is split, operation and maintenance flow and model parameters are difficult to dynamically iterate, and long-term operation requirements of an adaptive unit are insufficient. Therefore, an intelligent operation and maintenance method and system integrating multiple technical advantages, diagnosis precision, data safety and closed-loop optimization and integrating system architecture are needed, the bottleneck of the prior art is solved, and the operation and maintenance of a thermal power plant are promoted to develop to a more efficient, safe and intelligent direction. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides a fault intelligent diagnosis type power operation and maintenance method and system for a thermal power plant, which are used for solving the problems in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: on one hand, the invention provides an intelligent fault diagnosis type electric power operation and maintenance system of a thermal power plant, which adopts an edge-cloud-field three-layer cooperative architecture and comprises a sensing module, an edge processing module, a cloud core module, a field execution module and an optimization iteration module, wherein the modules are in redundant connection with an optical fiber network through a 5G industrial private network; The sensing module consists of a multi-type sensor cluster, a log acquisition unit, an infrared thermal imaging unit and a voiceprint acquisition unit and is used for acquiring structured and unstructured data of the whole set; The edge processing module adopts an industrial grade edge gateway, deploys a differential preprocessing unit, a space-time alignment unit and a federal learning encryption unit, and is used for data processing, space-time alignment and encryption hierarchical transmission; The cloud core module is based on a server cluster and comprises a three-level digital twin model unit, a three-mode fusion diagnosis engine, a fault propagation rule base unit, a blockchain evidence storage unit and a big data analysis unit, and is used for model construction and updating, fault diagnosis and prejudgment, and data evidence storage and analysis; The field execution module consists of a mobile terminal, an execution mechanism and redundant equipment and is used for receiving a work order, assisting in field treatment and switching equipment; The optimization iteration module is used for deploying a reinforcement learning unit and an operation and maintenance configuration optimization unit and is used for model iteration and operation and maintenance configuration optimization. On the other hand, the invention also provides a fault intelligent diagnosis type thermal power plant power operation and maintenance method, which is based on the system, and realizes the pre-judgment, diagnosis, treatment and optimization of t