A multi-dimensional intelligent inspection method and system for a new energy station

Abstract

The application relates to the technical field of intelligent inspection of new energy stations, and discloses a multi-dimensional intelligent inspection method and system for a new energy station, which collects data by using sensors, video equipment and remote acquisition devices, combines an image recognition algorithm in the field of artificial intelligence, and realizes centralized monitoring of power generation equipment and auxiliary control equipment in a power generation station; meanwhile, the method supports the free combination of multi-dimensional inspection modes such as routine inspection, special inspection, special inspection and self-defined inspection, realizes comprehensive verification of inspection results, and from the perspective of meeting the accuracy requirement of single data source, the online monitoring inspection and real-time data inspection method have lower costs than the video inspection, so that the number of expensive equipment such as cameras, unmanned aerial vehicles and track machines is reduced, hardware costs are further saved while the inspection work is completed, and finally, the goal of few-person and unmanned guarding of the new energy station is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology for new energy power stations, specifically a multi-dimensional intelligent inspection method and system for new energy power stations. Background Technology

[0002] Currently, the demand for reduced or unmanned operation of new energy power stations is becoming increasingly prominent. In the field of intelligent inspection of new energy power stations, most current work is carried out using manual inspections and single intelligent monitoring equipment. Manual inspections are inefficient and have a high rate of missed inspections, while the reliability of inspection results from single intelligent equipment cannot be guaranteed. Overall, the work still relies on manual review and cannot achieve the goal of reduced or unmanned operation.

[0003] In the field of new energy power generation, to ensure the efficient operation of power generation equipment, it is necessary to regularly monitor the operating status of equipment and components and related alarms. Therefore, each power generation group has taken the lead in formulating management methods for the operation and maintenance of power plant equipment. However, the current management methods still face many challenges in implementation: inconsistent inspection standards lead to uneven implementation results; paper-based records make it difficult to achieve data traceability and analysis; delayed fault warnings often miss the best handling window; and management experience among different power generation groups is difficult to share and reuse effectively. Traditional methods relying on manual inspection are not only labor-intensive but also lack comprehensive coverage of the inspected equipment and cannot meet the needs of real-time inspection. In addition, manual inspection is limited by the varying skill levels of personnel, resulting in insufficient ability to identify hidden defects, and the quality of inspection is difficult to guarantee in complex environments such as inclement weather and nighttime, further exacerbating equipment safety hazards.

[0004] The multi-dimensional intelligent inspection proposed in this invention utilizes various intelligent devices (sensors, video equipment, remote acquisition devices, etc.) and combines them with image recognition algorithms in the field of artificial intelligence to achieve centralized monitoring of power generation equipment and auxiliary control equipment in power plants. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-dimensional intelligent inspection method for new energy power stations. It has the advantages of multi-source data cross-validation, adaptive configuration of differentiated strategies, flexible combination of four types of inspections, hierarchical weighted fusion and automatic review linkage for low confidence, closed-loop management of results and automatic report generation. It solves the problems of low efficiency and incomplete coverage of manual inspections, insufficient reliability of single intelligent devices and reliance on manual review, poor timeliness and difficulty in ensuring compliance in complex environments, inconsistent standards and difficulty in data traceability, and high cost of video equipment.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional intelligent inspection method for new energy power stations, comprising the following steps: S1. Basic Data Collection: Collect a list of basic information for all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. Simultaneously, calculate the station's total annual designed power generation and, based on historical fault data, calculate the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; S2. Equipment Classification and Listing: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are classified into three categories: A, B, and C, forming a clear equipment classification list. S3. Differentiated configuration of inspection methods: Define the cost coefficient, reliability score and applicable scenario indicators of inspection methods for three types of equipment (A, B and C), establish a quantitative evaluation system for inspection methods, form a unified evaluation standard, and provide a basis for equipment classification and configuration. S4. Inspection Combination Scheme Generation: Based on the differentiated configuration results of inspection methods, generate inspection combination schemes for each level of equipment and output the "Station Inspection Method Configuration List" to complete the configuration phase closed loop; S5. Multi-source data access and acquisition: Connect the data channels of the three types of equipment inspection methods (A, B, and C), clean up invalid data, establish a standardized data format, and convert data from different sources into a unified format; S6. Initial Fault Assessment: Feature extraction and threshold comparison are performed on the standardized inspection data to complete the initial fault assessment for each inspection method, and the inspection results for each device are output. ; S7. Comprehensive Verification of Multi-Source Inspection Results: Calculate a weighted comprehensive score based on the weight of each inspection method and the results of a single inspection method. It performs a three-level result determination based on thresholds: no fault, divergence, and fault, and finally outputs a reliable equipment inspection result. S8. Manual review and result calibration: Weighted comprehensive scoring Results that are deemed to be in disagreement are manually reviewed to calibrate the final results and dynamically adjust the weighting of inspection methods. S9. Inspection Report Generation and Operation and Maintenance Application: Based on the execution of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task completion status, fault equipment information and manual review status.

[0007] Preferably, the annual power generation loss caused by the failure of a single device in S1 is... The calculation formula is:

[0008] In the formula, This indicates the unit's rated output affected by equipment failure. This indicates the average downtime / load reduction duration due to equipment failure. Indicates the fault compensation efficiency of backup power supply / equipment; The influence weight of power generation of a single device :

[0009] In the formula, This represents the theoretical total annual power generation of a new energy power plant.

[0010] Preferably, the equipment classification and inventory division process in S2 is as follows: S2.1, Classification based on the influence weight of power generation: Based on the influence weight of power generation of a single device calculated in S1. Classification is based on the following thresholds: (1) When the power generation of a single device affects the weight When the rate is ≥10%, it is classified as Class A critical equipment. When Class A equipment fails, it directly leads to unit shutdown or a significant loss of power generation. (2) When 3% ≤ the power generation impact weight of a single device When the efficiency is less than 10%, it is classified as Class B, which is a more important equipment. When Class B equipment fails, it will cause the unit to operate at reduced load or affect the power generation efficiency. (3) When the power generation of a single device affects the weight When the power generation is less than 3%, it is classified as Class C auxiliary equipment. When Class C equipment fails, it does not directly affect the power generation, but only the monitoring / regulation function. S2.2 Secondary verification based on the impact on power generation safety: Simultaneously perform secondary verification according to the degree of impact of equipment damage on power generation safety. For equipment with significant safety hazards, forcibly upgrade its classification by one level. S2.3 Output Classification List: Output the "Site Equipment Classification List", which clearly defines the equipment classification of each piece of equipment, the impact of failure, and the recommended inspection methods, thus completing the closed loop of the classification stage.

[0011] Preferably, the inspection combination scheme in S4 includes: (1) Class A equipment inspection configuration: Simultaneously configure three types of inspection methods: production data inspection, online monitoring inspection, and video inspection, to achieve cross-verification of three data sources, and increase the inspection frequency from once a day to once every 4 hours under special working conditions; (2) Class B equipment inspection configuration: The basic configuration consists of two types of inspection methods: production data inspection and online monitoring inspection. When visual inspection of towers or isolating switches is carried out, video inspection will be added. The inspection frequency is set to once a day, and under special working conditions, it will be increased to once every 8 hours. (3) Class C equipment inspection configuration: video inspection is preferred or only production data inspection is configured. The inspection frequency is once a week. Special inspections are triggered as needed under special working conditions.

[0012] Preferably, the S5 multi-source data acquisition: (1) Production data acquisition: Equipment operation data is acquired in real time through the centralized control system interface, and the data acquisition frequency is 1 time / minute; (2) Online monitoring data acquisition: Data from the online monitoring device is acquired through a dedicated communication interface at a frequency of once per hour; (3) Video inspection data collection: The equipment appearance images are captured by fixed cameras, drones and track machines along a preset route. The data collection frequency is synchronized with the inspection frequency, and defects are automatically identified by image recognition algorithms.

[0013] Preferably, the initial fault judgment rule for single-mode inspection in S6 is as follows: (1) Production data inspection: Compare real-time data with threshold ranges; if the data exceeds the threshold, it is considered abnormal. =0; otherwise, it is considered normal. =1; (2) Online monitoring and inspection: Compare the monitoring data with the early warning threshold; when the early warning conditions are met, it is determined to be abnormal. =0; otherwise, it is considered normal. =1; (3) Video inspection: The presence of appearance defects is determined by image recognition algorithms. When a defect is identified, it is judged as an anomaly. =0; otherwise, it is considered normal. =1; in, This represents the result of a single inspection of a certain piece of equipment using the j-th inspection method. The subscript j corresponds to different inspection methods: j=1: production data inspection; j=2: online monitoring inspection; j=3: video inspection.

[0014] Preferably, the multi-source inspection result comprehensive verification process in S7 is as follows: S7.1 For each piece of equipment, the number of inspection methods configured for it shall be determined accordingly. Calculate the weighted comprehensive score The calculation formula is as follows:

[0015] In the formula, This represents the base weight of the corresponding inspection method. This represents the result of a single inspection of a certain device using the j-th inspection method. S7.2 Result Judgment Rules: (1) Weighted composite score ≥0.7: The equipment is deemed to be fault-free, no further verification is required, and normal operation and maintenance procedures can be initiated; (2) 0.3 < weighted comprehensive score <0.7: This indicates a discrepancy in the multi-source data, triggering a manual review process; (3) Weighted composite score ≤0.3: The equipment is determined to be faulty, and the defect handling process is initiated directly, while a fault alarm message is generated.

[0016] Preferably, the manual verification and result calibration steps in S8 are as follows: S8.1. Maintenance personnel receive work orders with conflicting results, conduct manual judgment based on historical equipment data, on-site verification, and environmental factors, and output the final result. ; S8.2 Record the reasons for the disagreements to provide a basis for subsequent weight optimization; S8.3 When the manual review result deviates significantly from the comprehensive score result, the weight of the corresponding inspection method for that equipment will be automatically adjusted. S8.4 Finally, output the final inspection results for each device to complete the verification phase closed loop.

[0017] Preferably, the inspection report in S9 includes: (1) Basic inspection information: inspection cycle, equipment classification and inspection method configuration; (2) Core indicator data: equipment inspection coverage rate, task completion rate, manual review rate and cost saving rate; (3) Fault and alarm information: list of faulty equipment, fault classification, alarm distribution and handling suggestions; (4) Manual review status: list of divergent tasks, review results and optimization suggestions.

[0018] Preferably, in step S9, based on the inspection report, the following tasks are further performed: S9.1 Inspection plan iteration and optimization: Based on inspection data and review results, regularly review inspection data and continuously optimize hierarchical structure, configuration and algorithm; S9.2, Operation and Maintenance Process Integration and Unmanned Transition: Deeply integrate inspection results with operation and maintenance processes to gradually achieve less-staffed / unmanned operation.

[0019] A multi-dimensional intelligent inspection system for new energy power stations includes: Data acquisition module: Collects basic information on all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. It also calculates the station's total annual designed power generation and, based on historical fault data, calculates the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; Equipment classification and list division module: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are divided into three categories: A, B and C, forming a clear equipment classification list; The inspection method differentiation configuration module defines the cost coefficient, reliability score, and applicable scenario indicators for three types of equipment inspection methods (A, B, and C), establishes a quantitative evaluation system for inspection methods, forms a unified evaluation standard, and provides a basis for equipment classification and configuration. Inspection combination scheme generation module: Based on the differentiated configuration results of inspection methods, it generates inspection combination schemes for each level of equipment and outputs the "Site Inspection Method Configuration List", completing the configuration phase closed loop; Multi-source data access and acquisition module: It connects the data channels of three types of equipment inspection methods (A, B, and C), cleans up invalid data, establishes a standardized data format, and converts data from different sources into a unified format; Initial Fault Assessment Module: This module extracts features and compares thresholds on standardized inspection data to perform initial fault assessment for each inspection method and outputs the inspection results for each device. ; Multi-source inspection result comprehensive verification module: Calculates a weighted comprehensive score based on the weight of each inspection method and the inspection results of a single method. It performs a three-level result determination based on thresholds: no fault, divergence, and fault, and finally outputs a reliable equipment inspection result. Manual review and result calibration module: for weighted comprehensive scoring Results that are deemed to be in disagreement are manually reviewed to calibrate the final results and dynamically adjust the weighting of inspection methods. Inspection report generation and operation and maintenance application module: Based on the execution status of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task execution completion status, fault equipment information and manual review status.

[0020] Compared with existing technologies, this invention provides a multi-dimensional intelligent inspection method for new energy power stations, which has the following beneficial effects: This invention's multi-dimensional intelligent inspection system utilizes various intelligent devices and combines them with image recognition algorithms from the field of artificial intelligence to achieve centralized monitoring of power generation equipment and auxiliary control equipment in power plants. Specifically, it constructs a comprehensive data acquisition network for equipment status by deploying status sensing devices such as online water monitoring devices, online surge arrester monitoring devices, and online transformer oil chromatography monitoring devices. Simultaneously, it uses video inspection equipment such as high-definition cameras, infrared thermal imagers, drones, and tracked robots to achieve intelligent identification of equipment appearance and thermal defects. Relying on remote acquisition devices and SCADA systems, it acquires real-time electrical quantities and operating condition data of the equipment. The above three types of data sources are preprocessed locally through edge computing nodes and then converged to the central control center via a high-speed communication network, forming a centralized monitoring architecture that is cloud-edge-device collaborative.

[0021] This invention, through multi-source data access and acquisition, can meet the requirements for inspection accuracy. Among them, the online monitoring inspection and real-time production data inspection method is less expensive than video inspection, thereby avoiding the purchase of expensive equipment such as cameras, drones, and track machines, and further saving hardware costs while completing the inspection work.

[0022] 3. This invention supports the free combination of multi-dimensional inspection methods through routine inspections, special inspections, special inspections, and customized inspections, enabling comprehensive verification of inspection results. Routine inspections are automatically executed according to a preset cycle, covering daily equipment status monitoring; special inspections are triggered immediately under abnormal conditions such as extreme weather, power grid failures, and equipment tripping, enabling targeted troubleshooting; special inspections conduct in-depth inspections targeting specific equipment family defects or seasonal fault characteristics; and customized inspections allow maintenance personnel to flexibly configure inspection objects and methods according to temporary needs. The above four types of inspection tasks are optimized and configured through a unified task scheduling engine, which can effectively avoid the spatiotemporal conflicts of multi-source inspections, thereby ensuring the safe and orderly conduct of inspection operations.

[0023] 3. This invention establishes a hierarchical reliability evaluation system for different inspection methods, defining production data inspection as Level 1 reliability, online monitoring inspection as Level 2 reliability, and video inspection as Level 3 reliability. When inconsistencies occur in the results of multi-source inspections, the system can automatically calculate the comprehensive confidence level based on a weighted fusion algorithm, prioritizing the use of high-reliability data sources, and triggering manual review only when the confidence level is below the threshold or when there are conflicts in the results of important equipment. This minimizes human intervention while ensuring the accuracy of inspections, truly achieving the goal of minimal or even unmanned operation with machines as the main tool and human intervention as a supplement. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the operation of the method of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 A multi-dimensional intelligent inspection method for new energy power stations includes the following steps: S1. Basic Data Collection: Collect a list of basic information for all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. Simultaneously, calculate the station's total annual designed power generation and, based on historical fault data, calculate the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; S2. Equipment Classification and Listing: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are classified into three categories: A, B, and C, forming a clear equipment classification list to provide a basis for subsequent inspection strategy configuration. S3. Differentiated configuration of inspection methods: Define the cost coefficient, reliability score and applicable scenario indicators of inspection methods for three types of equipment (A, B and C), establish a quantitative evaluation system for inspection methods, form a unified evaluation standard, and provide a basis for equipment classification and configuration. S4. Inspection Combination Scheme Generation: Based on the differentiated configuration results of inspection methods, generate inspection combination schemes for each level of equipment and output the "Field Inspection Method Configuration List". This list clarifies the inspection method combination, frequency and data collection points for each piece of equipment, completing the configuration phase closed loop, thereby achieving accurate matching of inspection methods and balancing reliability and cost. S5. Multi-source data access and acquisition: Connect the data channels of the three types of equipment inspection methods (A, B, and C) to achieve unified data access. After cleaning invalid data (filtering outliers and missing values, marking failed data acquisition as invalid, triggering data re-acquisition or equipment fault alarm), establish a standardized data format to convert data from different sources into a unified format, thereby eliminating data heterogeneity. S6. Initial Fault Assessment: Feature extraction and threshold comparison are performed on the standardized inspection data to complete the initial fault assessment for each inspection method, and the inspection results for each device are output. This provides input for subsequent comprehensive verification; S7. Comprehensive Verification of Multi-Source Inspection Results: Calculate a weighted comprehensive score based on the weight of each inspection method and the results of a single inspection method. It performs three-level result determination based on thresholds: no fault, divergence, and fault, thereby solving the problem of inconsistency in multi-source data and outputting reliable equipment inspection results; S8. Manual review and result calibration: Weighted comprehensive scoring The result is considered divergent (0.3 < 0.3). Results with values ​​<0.7 are manually reviewed to calibrate the final results and dynamically adjust the weight of the inspection method. S9. Inspection Report Generation and Operation and Maintenance Application: Based on the execution of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task completion status, fault equipment information and manual review status.

[0027] Specifically, S1 includes basic data collection: annual power generation loss due to single equipment failure. (kW) The formula for calculating (h / year) is:

[0028] In the formula, Indicates the unit's rated output (kW) affected by equipment failure. This indicates the average downtime / load reduction duration due to equipment failure (h / time × average number of failures per year). This indicates the fault compensation efficiency of the backup power supply / equipment (0 if there is no backup). The influence weight of power generation of a single device :

[0029] In the formula, This represents the annual theoretical total power generation of new energy power plants (kW). h kWh).

[0030] The advantage is that it calculates the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit This provides an objective data basis for equipment classification, thereby avoiding subjective judgment and enabling subsequent inspection resources to be accurately directed to high-impact equipment.

[0031] Specifically, the equipment classification and inventory division process in S2: S2.1, Classification based on the influence weight of power generation: Based on the influence weight of power generation of a single device calculated in S1. Classification is based on the following thresholds: (1) When the power generation of a single device affects the weight When the rate is ≥10%, it is classified as Class A (important equipment). When Class A equipment fails, it will directly lead to unit shutdown or a significant loss of power generation. Specifically, it includes blades, main transformer, SVG, generator and grounding transformer. (2) When 3% ≤ the influence weight of power generation of a single device When the efficiency is less than 10%, it is classified as Category B (more important equipment). When Category B equipment fails, it will cause the unit to operate at reduced load or affect the power generation efficiency. Specifically, it includes the main shaft, gearbox, cooling system, bus PT, tower and isolating switch. (3) When the power generation of a single device affects the weight When the power generation is less than 3%, it is classified as Category C (auxiliary equipment). When Category C equipment fails, it does not directly affect the power generation, but only the monitoring / regulation function. Specifically, it includes protection pressure plates, synchronous condensers, AGC / AVC devices and overhead lines. S2.2 Secondary verification based on the impact on power generation safety: Simultaneously perform secondary verification according to the degree of impact of equipment damage on power generation safety. For equipment with significant safety hazards, forcibly upgrade its classification by one level (e.g., upgrade critical protection devices from Class C to Class B) to ensure that safety risks are fully covered. S2.3 Output Classification List: Output the "Site Equipment Classification List", which clearly defines the equipment classification (A / B / C), fault impact description and recommended inspection method for each piece of equipment. At this point, the equipment classification judgment and list division stage is completed (the classification stage is closed).

[0032] The advantages are: by adopting a dual mechanism that combines the weight threshold classification of power generation impact with the secondary verification of power generation safety impact, both economic efficiency and safety can be taken into account, making the classification of A / B / C class equipment more scientific and reasonable, thus providing a reliable basis for differentiated inspection strategies.

[0033] Specifically, the inspection methods in S3 are configured differently: (1) The relevant parameters for the three types of inspection methods are shown in Table 1 below: Table 1

[0034] (2) Define cost constraints: When the hardware cost of a single site inspection does not exceed 1% of the total equipment investment, the low-cost method should be given priority to meet the inspection needs; The advantages are: by establishing a quantitative evaluation system for inspection methods that includes reliability level, basic weight, cost and data source, the cost constraints (hardware cost ≤ 1% of total investment) are clarified, thereby providing a standardized basis for prioritizing the use of low-cost inspection methods (online monitoring, production data) while ensuring reliability.

[0035] Specifically, the inspection combination scheme in S4 includes: (1) Class A equipment inspection configuration: Simultaneously configure three types of inspection methods: production data inspection, online monitoring inspection, and video inspection, to achieve cross-verification of three data sources. Under special working conditions (such as typhoons, high temperatures, and periods of high incidence of equipment defects), the inspection frequency is increased from once a day to once every 4 hours. (2) Class B equipment inspection configuration: The basic configuration consists of two types of inspection methods: production data inspection and online monitoring inspection. When visual inspection of towers or isolating switches is carried out, video inspection will be added. The inspection frequency is set to once a day. Under special working conditions (such as strong winds and dust, icing, sudden load changes, and equipment warning periods), it will be increased to once every 8 hours. (3) Class C equipment inspection configuration: Prioritize video inspection (using existing surveillance cameras), or only configure production data inspection (such as protection pressure plate, status monitoring of synchronous condenser), the inspection frequency is once a week, and special inspections are triggered as needed (operation and maintenance judgment, scheduling instructions, adjacent faults, etc.) under special working conditions (such as system disturbance, after protection action, telemetry data jump, equipment self-inspection alarm).

[0036] The advantages are: by generating differentiated inspection combination schemes for three types of equipment (A, B, and C), precise matching is achieved; the three types of data sources are cross-validated for type A equipment, type B equipment is mainly carried out in a low-cost manner, and type C equipment prioritizes the use of existing cameras or production data; at the same time, it supports dynamic encryption frequency under special working conditions. This combination scheme can avoid purchasing a large number of expensive video equipment from the source, significantly saving hardware costs, thereby making online monitoring and production data inspection costs lower.

[0037] Specifically, S5 multi-source data acquisition: (1) Production data acquisition: Real-time acquisition of equipment operation data, including main transformer oil temperature, generator speed and bus voltage, through the centralized control system interface, with a data acquisition frequency of 1 time / minute; (2) Online monitoring data acquisition: Data from the online monitoring device is acquired through a dedicated communication interface, including the gas content of the main transformer oil chromatograph, the leakage current of the surge arrester and the trace moisture content of SF6 gas. The data acquisition frequency is 1 time / hour. (3) Video inspection data collection: The equipment appearance images are captured by fixed cameras, drones and track machines along a preset route. The data collection frequency is synchronized with the inspection frequency. At the same time, the image recognition algorithm is used to automatically identify defects (such as component corrosion, loose bolts, ice on the line, etc.).

[0038] The advantages are: The multi-dimensional intelligent inspection of this invention deploys status sensing devices such as micro-water online monitoring, surge arrester online monitoring, and oil chromatography, combined with video inspection equipment such as high-definition cameras, infrared thermal imaging, drones and track robots, and relies on remote acquisition devices and SCADA systems to build a comprehensive data acquisition network for equipment status. The above three types of data sources are preprocessed locally by edge computing nodes and then converged to the central control center through a high-speed communication network, forming a centralized monitoring architecture of cloud-edge-device collaboration, and finally realizing comprehensive, real-time and intelligent monitoring of power generation equipment and auxiliary control equipment.

[0039] Specifically, the initial fault diagnosis rules for single-mode inspection in S6 are as follows: (1) Production data inspection: Compare real-time data with threshold ranges; if the data exceeds the threshold, it is considered abnormal. =0; otherwise, it is considered normal. =1; (2) Online monitoring and inspection: Compare the monitoring data with the early warning threshold; when the early warning conditions are met, it is determined to be abnormal. =0; otherwise, it is considered normal. =1; (3) Video inspection: The presence of appearance defects is determined by image recognition algorithms. When a defect is identified, it is judged as an anomaly. =0; otherwise, it is considered normal. =1; in, This represents the result of a single inspection of a certain piece of equipment using the j-th inspection method. The subscript j corresponds to different inspection methods, for example: j=1: production data inspection; j=2: online monitoring inspection; j=3: video inspection.

[0040] The advantages are: by adopting a fault initial judgment rule that combines threshold comparison and image recognition, the rule is simple and efficient, and can quickly output single-mode inspection results. This provides reliable input for multi-source fusion verification, facilitating engineering implementation and real-time processing.

[0041] Specifically, the comprehensive verification process of multi-source inspection results in S7: S7.1 For each piece of equipment, the number of inspection methods configured for it shall be determined accordingly. Calculate the weighted comprehensive score The calculation formula is as follows:

[0042] In the formula, This indicates the basic weight of the corresponding inspection method (production data 0.6, online monitoring 0.3, video inspection 0.1). This represents the result of a single inspection of a certain device using the j-th inspection method. S7.2, Result Judgment Rule: (1) Weighted comprehensive score ≥0.7: It is determined that the equipment has no fault, no review is required, and the process enters the normal operation and maintenance process; (2) 0.3 < weighted comprehensive score < 0.7: It is determined that there is inconsistency among multi-source data, and the manual review process is triggered; (3) Weighted comprehensive score ≤ 0.3: It is determined that the equipment has a fault, the process directly enters the defect processing process, and fault alarm information is generated at the same time.

[0043] The advantage is that a hierarchical reliability evaluation system (production data for level 1, online monitoring for level 2, video inspection for level 3) is established through weighted comprehensive scoring and three-level threshold judgment (no fault, inconsistency, fault). When multi-source results are inconsistent, the system automatically calculates the comprehensive confidence degree based on the weighted fusion algorithm, gives priority to high-reliability data sources, and only triggers manual review in the inconsistency interval (0.3<S<0.7), thereby minimizing manual intervention on the premise of ensuring accuracy, and truly realizing the goal of less human participation with machines as the main part and humans as the auxiliary part.

[0044] Specifically, the steps of manual review and result calibration in S8 are as follows: S8.1, operation and maintenance personnel receive the work order of inconsistent results, conduct manual judgment by combining equipment historical data, on-site verification and environmental factors (such as weather, load change), and output the final result (1 represents normal, 0 represents fault); S8.2, record the cause of inconsistency (such as misjudgment of video inspection, fluctuation of online monitoring data, temporary abnormality of production data, etc.), to provide a basis for subsequent weight optimization; S8.3, when the deviation between the manual review result and the comprehensive scoring result is large, automatically adjust the weight of the corresponding inspection mode of the equipment: when the misjudgment rate of a certain mode exceeds 20%, reduce the weight by 0.05, and increase the weight of other modes in the same group by an equal amount, keeping the sum of weights as 1; S8.4, finally output the final inspection result of each equipment, and complete the closed loop of the verification stage.

[0045] The advantage is that by introducing a closed-loop mechanism for manual review, the cause of inconsistency can be recorded in real time and the weight can be adjusted dynamically. When the misjudgment rate of a certain mode exceeds 20%, the weight is automatically reduced by 0.05, and the weight of other modes in the same group is increased by an equal amount, keeping the sum of weights as 1. This mechanism endows the method of the present invention with adaptive optimization capability, and while continuously improving the accuracy of comprehensive verification, it can also ensure that manual review only focuses on a few uncertain results, thereby further reducing dependence on manual work.

[0046] Specifically, the inspection report in S9 includes: (1) Basic inspection information: inspection cycle, equipment classification and inspection method configuration; (2) Core indicator data: equipment inspection coverage rate, task completion rate, manual review rate and cost saving rate; (3) Fault and alarm information: list of faulty equipment, fault classification, alarm distribution and handling suggestions; (4) Manual review status: list of divergent tasks, review results and optimization suggestions.

[0047] Specifically, in S9, based on the inspection report, the following tasks are further performed: S9.1 Inspection Plan Iterative Optimization: Based on inspection data and review results, conduct regular (quarterly) reviews of inspection data and continuously optimize the hierarchical structure, configuration, and algorithm, as detailed below: (1) Adjust equipment classification: Update the power generation impact weight of a single piece of equipment according to the actual impact of equipment failure on power generation. Reclassify equipment into A / B / C categories; (2) Optimize the configuration of inspection methods: For equipment with high verification rate and high misjudgment rate, adjust the combination of inspection methods (such as adding online monitoring equipment and adjusting the frequency of video inspection). (3) Optimize reliability weights: Adjust the base weights based on the actual accuracy of various inspection methods. This will improve the accuracy of comprehensive verification. (4) Optimize image recognition algorithm: Based on the review results of video inspection, update the defect recognition model to reduce the false judgment rate; S9.2, Operation and Maintenance Process Integration and Unmanned Transition: Deeply integrate inspection results with operation and maintenance processes to gradually achieve minimal or no human intervention, as detailed below: (1) Faulty equipment directly pushes defect handling work orders, automatically associates equipment historical fault records and maintenance manuals, generates maintenance suggestions, and reduces manual decision-making costs; (2) Reduce the frequency of manual inspections: When the manual review rate is consistently below 5% and the equipment failure rate is stable, gradually reduce the frequency of on-site manual inspections and prioritize the use of intelligent inspection methods. (3) Establish an unattended emergency mechanism: Set up a dual alarm mechanism for Class A equipment. When multiple data sources simultaneously determine a fault, automatically trigger the shutdown protection to ensure equipment safety.

[0048] The advantages are: by generating multi-dimensional reports that include basic inspection information, core indicators, fault alarms, and manual review, and by regularly (quarterly) reviewing the reports, the equipment classification, inspection method combination, reliability weight, and image recognition algorithm can be dynamically adjusted to form a continuous improvement closed loop; at the same time, the inspection results are deeply integrated with the operation and maintenance process. The above mechanism effectively promotes the transformation from manual regular inspection to intelligent on-demand inspection + unattended operation mode.

[0049] In summary, the method of this invention utilizes various intelligent devices (sensors, video equipment, remote data acquisition devices, etc.) combined with image recognition algorithms in the field of artificial intelligence to achieve centralized monitoring of power generation equipment and auxiliary control equipment in power plants. Furthermore, the proposed routine inspections, special inspections, dedicated inspections, and customized inspections support the free combination of multi-dimensional inspection methods, enabling comprehensive verification of inspection results. In addition, from the perspective of meeting inspection accuracy requirements from a single data source, online monitoring and real-time production data inspection methods are more cost-effective than video inspections, thereby reducing the procurement investment in expensive equipment such as cameras, drones, and track machines. This further saves hardware costs while completing the inspection work, ultimately achieving the goal of reduced or unmanned operation of new energy power plants.

[0050] In another embodiment of the present invention, a multi-dimensional intelligent inspection system for new energy power stations is provided, comprising: Data acquisition module: Collects basic information on all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. It also calculates the station's total annual designed power generation and, based on historical fault data, calculates the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; Equipment classification and list division module: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are divided into three categories: A, B and C, forming a clear equipment classification list; The inspection method differentiation configuration module defines the cost coefficient, reliability score, and applicable scenario indicators for three types of equipment inspection methods (A, B, and C), establishes a quantitative evaluation system for inspection methods, forms a unified evaluation standard, and provides a basis for equipment classification and configuration. Inspection combination scheme generation module: Based on the differentiated configuration results of inspection methods, it generates inspection combination schemes for each level of equipment and outputs the "Site Inspection Method Configuration List", completing the configuration phase closed loop; Multi-source data access and acquisition module: It connects the data channels of three types of equipment inspection methods (A, B, and C), cleans up invalid data, establishes a standardized data format, and converts data from different sources into a unified format; Initial Fault Assessment Module: This module extracts features and compares thresholds on standardized inspection data to perform initial fault assessment for each inspection method and outputs the inspection results for each device. ; Multi-source inspection result comprehensive verification module: Calculates a weighted comprehensive score based on the weight of each inspection method and the inspection results of a single method. It performs a three-level result determination based on thresholds: no fault, divergence, and fault, and finally outputs a reliable equipment inspection result. Manual review and result calibration module: for weighted comprehensive scoring Results that are deemed to be in disagreement are manually reviewed to calibrate the final results and dynamically adjust the weighting of inspection methods. Inspection report generation and operation and maintenance application module: Based on the execution status of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task execution completion status, fault equipment information and manual review status.

[0051] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional intelligent inspection method for new energy power stations, characterized in that, Includes the following steps: S1. Basic Data Collection: Collect a list of basic information for all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. Simultaneously, calculate the station's total annual designed power generation and, based on historical fault data, calculate the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; S2. Equipment Classification and Listing: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are classified into three categories: A, B, and C, forming a clear equipment classification list. S3. Differentiated configuration of inspection methods: Define the cost coefficient, reliability score and applicable scenario indicators of inspection methods for three types of equipment (A, B and C), establish a quantitative evaluation system for inspection methods, form a unified evaluation standard, and provide a basis for equipment classification and configuration. S4. Inspection Combination Scheme Generation: Based on the differentiated configuration results of inspection methods, generate inspection combination schemes for each level of equipment and output the "Station Inspection Method Configuration List" to complete the configuration phase closed loop; S5. Multi-source data access and acquisition: Connect the data channels of the three types of equipment inspection methods (A, B, and C), clean up invalid data, establish a standardized data format, and convert data from different sources into a unified format; S6. Initial Fault Assessment: Feature extraction and threshold comparison are performed on the standardized inspection data to complete the initial fault assessment for each inspection method, and the inspection results for each device are output. ; S7. Comprehensive Verification of Multi-Source Inspection Results: Calculate a weighted comprehensive score based on the weight of each inspection method and the results of a single inspection method. It performs a three-level result determination based on thresholds: no fault, divergence, and fault, and finally outputs a reliable equipment inspection result. S8. Manual review and result calibration: Weighted comprehensive scoring Results that are deemed to be in disagreement are manually reviewed to calibrate the final results and dynamically adjust the weighting of inspection methods. S9. Inspection Report Generation and Operation and Maintenance Application: Based on the execution of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task completion status, fault equipment information and manual review status.

2. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The annual power generation loss caused by the failure of a single device in S1 The calculation formula is: In the formula, This indicates the unit's rated output affected by equipment failure. This indicates the average downtime / load reduction duration due to equipment failure. Indicates the fault compensation efficiency of backup power supply / equipment; The influence weight of power generation of a single device : In the formula, This represents the theoretical total annual power generation of a new energy power plant.

3. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The equipment classification and inventory division process in S2 is as follows: S2.1, Classification based on the influence weight of power generation: Based on the influence weight of power generation of a single device calculated in S1. Classification is based on the following thresholds: (1) When the power generation of a single device affects the weight When the rate is ≥10%, it is classified as Class A critical equipment. When Class A equipment fails, it directly leads to unit shutdown or a significant loss of power generation. (2) When 3% ≤ the influence weight of power generation of a single device When the efficiency is less than 10%, it is classified as Class B, which is a more important type of equipment. When Class B equipment fails, it will cause the unit to operate at reduced load or affect the power generation efficiency. (3) When the power generation of a single device affects the weight When the power generation is less than 3%, it is classified as Class C auxiliary equipment. When Class C equipment fails, it does not directly affect the power generation, but only the monitoring / regulation function. S2.2 Secondary verification based on the impact on power generation safety: Simultaneously perform secondary verification according to the degree of impact of equipment damage on power generation safety. For equipment with significant safety hazards, forcibly upgrade its classification by one level. S2.3 Output Classification List: Output the "Site Equipment Classification List", which clearly defines the equipment classification of each piece of equipment, the impact of failure, and the recommended inspection methods, thus completing the closed loop of the classification stage.

4. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The inspection combination scheme in S4 includes: (1) Class A equipment inspection configuration: Simultaneously configure three types of inspection methods: production data inspection, online monitoring inspection, and video inspection, to achieve cross-verification of three data sources, and increase the inspection frequency from once a day to once every 4 hours under special working conditions; (2) Class B equipment inspection configuration: The basic configuration consists of two types of inspection methods: production data inspection and online monitoring inspection. When visual inspection of towers or isolating switches is carried out, video inspection will be added. The inspection frequency is set to once a day, and under special working conditions, it will be increased to once every 8 hours. (3) Class C equipment inspection configuration: video inspection is preferred or only production data inspection is configured. The inspection frequency is once a week. Special inspections are triggered as needed under special working conditions.

5. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The S5 multi-source data acquisition: (1) Production data acquisition: Equipment operation data is acquired in real time through the centralized control system interface, and the data acquisition frequency is 1 time / minute; (2) Online monitoring data acquisition: Data from the online monitoring device is acquired through a dedicated communication interface at a frequency of once per hour; (3) Video inspection data collection: The equipment appearance images are captured by fixed cameras, drones and track machines along a preset route. The data collection frequency is synchronized with the inspection frequency, and defects are automatically identified by image recognition algorithms.

6. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The initial fault judgment rules for single-mode inspection in S6 are as follows: (1) Production data inspection: Compare real-time data with threshold ranges; if the data exceeds the threshold, it is considered abnormal. =0; Otherwise, it is considered normal. =1; (2) Online monitoring and inspection: Compare the monitoring data with the early warning threshold; when the early warning conditions are met, it is determined to be abnormal. =0; Otherwise, it is considered normal. =1; (3) Video inspection: The presence of appearance defects is determined by image recognition algorithms. When a defect is identified, it is judged as an anomaly. =0; Otherwise, it is considered normal. =1; in, This represents the result of a single inspection of a certain piece of equipment using the j-th inspection method. The subscript j corresponds to different inspection methods: j=1: production data inspection; j=2: online monitoring inspection; j=3: video inspection.

7. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The comprehensive verification process of multi-source inspection results in S7: S7.1 For each piece of equipment, the number of inspection methods configured for it shall be determined accordingly. Calculate the weighted comprehensive score The calculation formula is as follows: In the formula, This represents the base weight of the corresponding inspection method. This represents the result of a single inspection of a certain piece of equipment using the j-th inspection method. S7.2 Result Judgment Rules: (1) Weighted composite score ≥0.7: The equipment is deemed to be fault-free, no further verification is required, and normal operation and maintenance procedures can be initiated; (2) 0.3 < weighted comprehensive score <0.7: This indicates a discrepancy in the multi-source data, triggering a manual review process; (3) Weighted composite score ≤0.3: The equipment is determined to be faulty, and the defect handling process is initiated directly, while a fault alarm message is generated.

8. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The manual review and result calibration steps in S8 are as follows: S8.1 Maintenance personnel receive work orders with conflicting results, conduct manual judgment based on historical equipment data, on-site verification, and environmental factors, and output the final result. ; S8.2 Record the reasons for the disagreements to provide a basis for subsequent weight optimization; S8.3 When the manual review result deviates significantly from the comprehensive score result, the weight of the corresponding inspection method for that equipment will be automatically adjusted. S8.4 Finally, output the final inspection results for each device to complete the verification phase closed loop.

9. The multi-dimensional intelligent inspection method for new energy power stations according to claim 1, characterized in that, The inspection report in S9 includes: (1) Basic inspection information: inspection cycle, equipment classification and inspection method configuration; (2) Core indicator data: equipment inspection coverage rate, task completion rate, manual review rate and cost saving rate; (3) Fault and alarm information: list of faulty equipment, fault classification, alarm distribution and handling suggestions; (4) Manual review status: list of conflicting tasks, review results and optimization suggestions; Based on the inspection report, the following tasks will be performed: S9.1 Inspection plan iteration and optimization: Based on inspection data and review results, regularly review inspection data and continuously optimize hierarchical structure, configuration and algorithm; S9.2, Operation and Maintenance Process Integration and Unmanned Transition: Deeply integrate inspection results with operation and maintenance processes to gradually achieve less-staffed / unmanned operation.

10. A multi-dimensional intelligent inspection system for new energy power stations, characterized in that, include: Data acquisition module: Collects basic information on all equipment in the new energy power station, including equipment name, model, installation location, function, historical fault records and fault handling time. It also calculates the station's total annual designed power generation and, based on historical fault data, calculates the annual power generation loss caused by the failure of a single piece of equipment. Weighting of power generation per unit ; Equipment classification and list division module: Based on the importance of their impact on power generation, the power generation equipment and auxiliary control equipment of new energy power plants are divided into three categories: A, B and C, forming a clear equipment classification list; The inspection method differentiation configuration module defines the cost coefficient, reliability score, and applicable scenario indicators for three types of equipment inspection methods (A, B, and C), establishes a quantitative evaluation system for inspection methods, forms a unified evaluation standard, and provides a basis for equipment classification and configuration. Inspection combination scheme generation module: Based on the differentiated configuration results of inspection methods, it generates inspection combination schemes for each level of equipment and outputs the "Site Inspection Method Configuration List", completing the configuration phase closed loop; Multi-source data access and acquisition module: It connects the data channels of three types of equipment inspection methods (A, B, and C), cleans up invalid data, establishes a standardized data format, and converts data from different sources into a unified format; Initial Fault Assessment Module: This module extracts features and compares thresholds on standardized inspection data to perform initial fault assessment for each inspection method and outputs the inspection results for each device. ; Multi-source inspection result comprehensive verification module: Calculates a weighted comprehensive score based on the weight of each inspection method and the inspection results of a single method. It performs a three-level result determination based on thresholds: no fault, divergence, and fault, and finally outputs a reliable equipment inspection result. Manual review and result calibration module: for weighted comprehensive scoring Results that are deemed to be in disagreement are manually reviewed to calibrate the final results and dynamically adjust the weighting of inspection methods. Inspection report generation and operation and maintenance application module: Based on the execution status of inspection tasks, inspection reports are generated periodically or manually, and the solutions are iterated and optimized and the operation and maintenance are connected based on the reports. The generated reports include equipment inspection coverage, task execution completion status, fault equipment information and manual review status.

Images