An artificial intelligence-based power plant inspection result analysis method and system

Abstract

The application discloses a power plant inspection result analysis method and system based on artificial intelligence, relates to the technical field of power plant management, and collects the space positions of inspectors to generate a track stay time vector, quantifies the actual movement path and stay area of the inspectors in the power plant, combines the attention degree and stay condition in the corresponding area, improves the integrity of the inspection behavior data through the space track and the stay time vector, reflects the inspection quality through a visual attention matrix, associates the inspection behavior with the equipment state, quantizes the abnormal tendency through an operation habit vector, provides data support for comprehensively evaluating the inspection behavior quality and early warning potential risks, and multi-dimensionally fuses three types of data, i.e., the space track, the visual attention and the operation habit, to analyze the inspection result from multiple aspects, overcome the limitation of single dimension analysis, and effectively improve the accuracy and robustness of the inspection abnormality detection.

Description

Technical Field

[0001] This invention relates to the field of power plant management technology, and in particular to a method and system for analyzing power plant inspection results based on artificial intelligence. Background Technology

[0002] Power plant inspections rely on a combination of manual inspections and some automated monitoring methods. Traditional manual inspections are usually based on scheduled on-site inspections along fixed routes. Inspectors need to carry handheld terminals or record sheets to conduct visual inspections, read instrument readings, listen to sounds, and perform simple operations on key equipment such as generator sets, transformers, and switchgear. If any abnormalities are found, they are then reported and handled manually. Some power plants have introduced automated sensing equipment such as video surveillance, infrared temperature measurement, and vibration sensors to achieve remote real-time monitoring and alarms for key areas. However, most automated systems operate independently and fail to form an effective linkage with inspection activities.

[0003] Power plant inspections are a crucial link in ensuring equipment safety and stable operation. Traditional inspections rely on human experience, which is inefficient and prone to overlooking potential hazards. The introduction of artificial intelligence technology has improved the automation and intelligence level of power plant inspections. By using artificial intelligence algorithms such as machine learning and deep learning, the behavior trajectory of inspection personnel can be analyzed to automatically identify whether the inspection is standardized and whether there are omissions or abnormal operations. However, it is unable to effectively capture and analyze implicit information such as the distribution of the inspection personnel's attention and operating habits, which leads to erroneous analysis results of the inspection results. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an artificial intelligence-based method for analyzing power plant inspection results to solve the problem of inaccurate inspection results.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] Firstly, the present invention provides a method for analyzing power plant inspection results based on artificial intelligence, including:

[0008] S1. Collect the three-dimensional spatial trajectory data and timestamps of the inspection personnel, and generate a sequence of trajectory points;

[0009] S2. Based on the collected trajectory point sequence, combined with head posture information, the position of the gaze focus is analyzed to generate the attention distribution matrix of the inspection personnel to the equipment components;

[0010] S3. Collect the operation event sequence of the inspection personnel, convert it into a symbol sequence based on the time order, extract the operation pattern, and generate an operation habit feature vector;

[0011] S4. Standardize and fuse the collected three-dimensional spatial trajectory data, timestamps, attention distribution matrix, and operation habit feature vectors to form inspection behavior feature vectors;

[0012] S5. Collect power plant equipment status parameters and environmental parameters, construct a multivariate nonlinear coupled model, dynamically estimate model parameters through a recursive algorithm, and generate predicted values ​​and residuals of equipment status.

[0013] S6. Based on the equipment status residual and the inspection behavior feature vector, perform anomaly judgment, and combine the causal reasoning mechanism to generate anomaly causes and propagation paths based on the equipment and environmental causal graph.

[0014] As a preferred embodiment of the power plant inspection result analysis method based on artificial intelligence described in this invention, in step S1, an ultra-wideband positioning device is used to collect the spatial position of the inspection personnel in real time to obtain three-dimensional coordinate data. The collected three-dimensional coordinate data is sorted according to the time sequence to form a trajectory point sequence. The trajectory point sequence is filtered to remove abnormal noise and obtain smooth trajectory data. The time interval between trajectory data is obtained based on the timestamp to obtain the trajectory dwell time vector. The trajectory dwell time vector is used to reflect the trajectory of the spatial motion state of the inspection.

[0015] As a preferred embodiment of the power plant inspection result analysis method based on artificial intelligence described in this invention, in step S2, a head posture sensor is used to collect head posture angle data of the inspection personnel, generating a posture angle sequence. Based on the posture angle sequence and the three-dimensional coordinate data collected by the ultra-wideband positioning device, a gaze direction vector is generated. A fixed gaze projection distance is set, and a correlation analysis is performed on the gaze direction vector and the three-dimensional coordinate data to generate a gaze focus position. The gaze focus position is mapped to a preset equipment area, and the cumulative gaze dwell time in each area is counted. An attention distribution vector is constructed based on the gaze dwell time. The attention distribution vector is used to reflect the degree of attention that the inspection personnel pay to key equipment.

[0016] As a preferred embodiment of the power plant inspection result analysis method based on artificial intelligence described in this invention, in step S3, operation events of inspection personnel are collected, the operation events include operation type and target equipment identifier, operation event sequence is generated based on operation events, the operation event sequence is encoded based on time order and converted into symbol sequence, correlation analysis is performed on symbol sequence based on frequent subsequence mining algorithm to identify high-frequency operation patterns and generate corresponding occurrence frequencies of operation patterns, and operation habit feature vector is constructed, the operation habit feature vector is used to reflect the operation rules of inspection personnel.

[0017] As a preferred embodiment of the artificial intelligence-based power plant inspection result analysis method described in this invention, in step S4, the trajectory dwell time vector, the attention distribution vector, and the operation habit feature vector obtained by the positioning acquisition module are normalized. Correlation analysis is then performed on the normalized trajectory dwell time vector, operation habit feature vector, and attention distribution vector to generate feature value B. The formula used is:

[0018] ,

[0019] in, , and These represent the normalized trajectory dwell time vector, attention distribution vector, and operation habit feature vector, respectively, with k used to index the elements in the operation habit feature vector. The norm is represented by T, the total inspection time is T, and B is the eigenvalue, which is used to reflect the comprehensive performance of multi-source behavioral data.

[0020] As a preferred embodiment of the artificial intelligence-based power plant inspection result analysis method of the present invention, in step S5, power plant equipment status parameters and environmental parameters are collected, and a status parameter vector is generated. and environmental parameter vector A nonlinear coupled model of Gaussian radial basis functions is constructed to generate predicted equipment conditions. The formula used is as follows:

[0021] ,

[0022] in, This is the weight matrix. The kernel center vector, B is the kernel width, and B is the eigenvalue. Here, M represents the number of Gaussian radial basis function kernels, and m is used to index the Gaussian radial basis function kernels. Based on the recursive least squares algorithm, the model parameters are analyzed to generate predicted equipment status values. and generate residuals The formula used is:

[0023] ,

[0024] in, For the state parameter vector, the predicted device state value The residual is used to reflect the device state parameter vector predicted at time t. Used to reflect the difference between the actual state of the equipment and the predicted state.

[0025] As a preferred embodiment of the artificial intelligence-based power plant inspection result analysis method of the present invention, wherein: in step S6, correlation analysis is performed on the residuals and behavioral feature values ​​to generate a comprehensive equipment anomaly score. The formula used is:

[0026] ,

[0027] Among them, the comprehensive anomaly score Used to determine the equipment risk index. and The parameters are used to adjust the nonlinear influence of behavioral characteristics on the anomaly score and the contribution of the integral term to the total anomaly score, respectively. T is the total inspection time, based on the comprehensive anomaly score. Determine if an anomaly exists, generate the cause and propagation path of the anomaly through a cause-effect graph of the device and environment, and output a structured anomaly report.

[0028] Secondly, this invention provides an artificial intelligence-based power plant inspection result analysis system, including:

[0029] The positioning and acquisition module is used to collect the three-dimensional spatial trajectory data and timestamps of the inspection personnel and generate a sequence of trajectory points.

[0030] The gaze capture module is used to analyze the gaze focus position based on the collected trajectory point sequence and head posture information, and generate an attention distribution matrix of the inspection personnel to the equipment components.

[0031] The operation coding module is used to collect the operation event sequence of inspection personnel, convert it into a symbol sequence based on time order, extract the operation pattern, and generate an operation habit feature vector.

[0032] The behavior fusion module is used to standardize and fuse the collected three-dimensional spatial trajectory data, timestamps, attention distribution matrix, and operation habit feature vectors to form inspection behavior feature vectors.

[0033] The state coupling module is used to collect power plant equipment state parameters and environmental parameters, construct a multivariate nonlinear adaptive coupling model, and dynamically estimate the model parameters through a recursive algorithm to obtain the predicted values ​​and residuals of the equipment state.

[0034] The anomaly analysis module is used to determine anomalies based on equipment status residuals and inspection behavior feature vectors. Combined with a causal reasoning mechanism, it generates anomaly causes and propagation paths based on the causal graph of equipment and environment.

[0035] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the power plant inspection result analysis method based on artificial intelligence as described in the first aspect of the present invention.

[0036] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the power plant inspection result analysis method based on artificial intelligence as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: By analyzing the inspection behavior of inspection personnel, the degree of abnormality of power plant equipment is obtained through coupling. Specifically, spatial location data of inspection personnel is collected to obtain three-dimensional coordinate data, which is then sorted by time to generate a trajectory dwell time vector. The actual movement path and dwell area of ​​inspection personnel within the power plant are quantified, and combined with the attention and dwell status in the corresponding areas, the completeness of inspection behavior data is improved through spatial trajectory and dwell time vector. The inspection quality is reflected through a visual attention matrix, which links inspection behavior with equipment status. The abnormal tendency is quantified through an operation habit vector, providing data support for comprehensively evaluating the quality of inspection behavior and warning of potential risks. The multi-dimensional fusion of spatial trajectory, visual attention, and operation habit data allows for multi-faceted analysis of inspection results, overcoming the limitations of single-dimensional analysis and effectively improving the accuracy and robustness of inspection anomaly detection. Correlation analysis is performed on residuals and behavioral feature values ​​to generate a comprehensive equipment anomaly score that can accurately determine equipment anomalies, thereby improving the early warning capability of power plant equipment inspection. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 A flowchart of a power plant inspection result analysis method based on artificial intelligence;

[0040] Figure 2 This is a flowchart illustrating the analysis of the attention distribution vector in this invention. Detailed Implementation

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0042] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0044] Reference Figures 1-2 This is one embodiment of the present invention, which provides a method for analyzing power plant inspection results based on artificial intelligence, including:

[0045] S1. Collect the three-dimensional spatial trajectory data and timestamps of the inspection personnel, and generate a sequence of trajectory points;

[0046] An ultra-wideband positioning device is used to collect the spatial position of the inspection personnel in real time and obtain three-dimensional coordinate data. The collected three-dimensional coordinate data is sorted according to the time series to form a trajectory point sequence. The trajectory point sequence is filtered to remove abnormal noise and obtain smooth trajectory data. The time interval between trajectory data is obtained based on the timestamp to obtain the trajectory dwell time vector. The trajectory dwell time vector is used to reflect the trajectory's spatial motion state during inspection.

[0047] Based on ultra-wideband positioning technology, the three-dimensional spatial coordinates of inspection personnel in the power plant environment are collected by an ultra-wideband positioning device. The sampling frequency is set to 10Hz. Ultra-wideband positioning technology has strong anti-interference and high accuracy, making it suitable for the power plant environment. The trajectory point sequence provides the spatial motion basis for the subsequent line-of-sight acquisition module.

[0048] When inspecting power plant equipment, inspection personnel will spend different amounts of time in different areas. By analyzing the inspection trajectory, it can be determined whether the inspection personnel are following the prescribed route and whether there are any missed inspections or abnormal stops. The inspection scenario is divided into N spatial sub-regions, with i as the index for each sub-region. The trajectory data of the inspection personnel is a time-series spatial coordinate. T represents the total inspection time, and the trajectory dwell time vector. The formula used to represent the cumulative time spent by the inspection personnel in each sub-region of the i-th region is:

[0049] ;

[0050] in, Let i be the spatial extent of the i-th spatial sub-region. This is an indicator function; it takes the value 1 if the condition is true, and 0 otherwise. It represents the trajectory dwell time vector. Used to reflect the inspection personnel's performance in the area Total dwell time is calculated by accumulating the dwell time of inspection personnel in each area to establish a spatial dwell time distribution. Longer dwell time usually means that the equipment in that area has received more attention or that the inspection process is more complex. Collecting this vector can reflect the movement path of inspection personnel and the key areas they stay in.

[0051] S2. Based on the collected trajectory point sequence, combined with head posture information, the position of the gaze focus is analyzed to generate the attention distribution matrix of the inspection personnel to the equipment components;

[0052] A head posture sensor is used to collect head posture angle data of the inspection personnel, generating a posture angle sequence. Based on the posture angle sequence and the three-dimensional coordinate data collected by the ultra-wideband positioning device, a gaze direction vector is generated. A fixed gaze projection distance is set, and correlation analysis is performed on the gaze direction vector and the three-dimensional coordinate data to generate the gaze focus position. The gaze focus position is mapped to a preset equipment area, and the cumulative gaze dwell time in each area is counted. An attention distribution vector is constructed based on the gaze dwell time. The attention distribution vector is used to reflect the degree of attention of the inspection personnel to the key equipment.

[0053] Step 2 is used to determine the actual location of the inspection personnel's gaze, i.e., which key parts they are inspecting. Therefore, it is also necessary to collect the direction of the inspection personnel's attention to determine whether they are focused on the corresponding equipment. Through the gaze data, it can be determined whether the inspection personnel are carefully inspecting the equipment and paying attention to key links, thereby improving the quality of the inspection assessment. It can also help to find problems such as lack of focus or omission of key points during the inspection. Subsequent behavior fusion analysis can improve the accuracy of anomaly identification.

[0054] The attention distribution vector A reflects the degree of visual focus of inspection personnel on different critical equipment areas, demonstrating the visual focus situation. It is calculated using head posture data, including pitch, yaw, and roll angles, combined with positioning data, to determine the inspection personnel's line-of-sight direction vector. Then, combined with the distance of the line of sight projection, determine the position of the focal point of the line of sight. The inspection area is divided into M key equipment areas. Calculate the cumulative fixation time of the inspection personnel's gaze on each equipment area, and define the j-th component of the attention distribution vector as:

[0055] ;

[0056] in, d represents the distance to be projected. For the j-th device region, the j-th component of the attention distribution vector Let denoted as the visual dwell time on the j-th equipment area. The visual focus of the inspector is an important indicator for judging whether the inspection is comprehensive. The attention distribution vector reflects the attention distribution on different equipment areas. The longer the attention time, the more thorough the key inspection and examination of that location.

[0057] S3. Collect the operation event sequence of the inspection personnel, convert it into a symbol sequence based on the time order, extract the operation pattern, and generate an operation habit feature vector;

[0058] The system collects operation events of inspection personnel, including operation type and target equipment identifier. Based on the operation events, it generates operation event sequences, encodes the operation event sequences according to time order, converts them into symbol sequences, performs correlation analysis on the symbol sequences based on frequent subsequence mining algorithm, identifies high-frequency operation patterns, generates the corresponding occurrence frequency of operation patterns, and constructs operation habit feature vectors. The operation habit feature vectors are used to reflect the operation patterns of inspection personnel.

[0059] The specific operations of the inspection personnel, such as switching equipment on and off and adjusting instruments, are compiled in chronological order. The operation patterns are analyzed. The inspection operations reflect the actual work content. Regular operations illustrate the work standards. Abnormal operations may indicate errors or potential accidents. By analyzing the operation sequence, frequent operation patterns can be identified, the work habits of the inspection personnel can be obtained, and potential risks can be analyzed.

[0060] The operation habit vector H is used to reflect the operational behavior patterns of inspection personnel, reflecting common operation modes and frequencies, and assisting in judging whether the operation is standardized, as well as the sequence of inspection operation events. Encode and statistically analyze frequent operation patterns. Each element of the inspection operation event sequence contains the operation type, timestamp, and target device. Encode the operation event sequence into a symbol sequence. Each element in the symbol sequence represents an operation type. Frequent subsequence mining algorithms are used to extract frequently occurring subsequence patterns from the symbol sequence. Count the frequency of each frequent pattern in the sequence. The formula for generating the operational habit vector H is as follows:

[0061] ;

[0062] Where K is the number of frequent operation modes, and frequency. Let be the frequency of occurrence of the k-th frequent operation pattern, representing the strength of the operation habit. For operation events, Symbol encoding for the corresponding operation, As a frequent subsequence pattern, by analyzing the order and frequency of operation events, it can reflect the operation pattern of the inspection personnel. Frequently occurring patterns indicate normal or standard operations, while abnormal or rare operations may indicate abnormalities or errors in the inspection process. After vectorization, it provides quantitative features for subsequent behavior fusion.

[0063] S4. Standardize and fuse the collected three-dimensional spatial trajectory data, timestamps, attention distribution matrix, and operation habit feature vectors to form inspection behavior feature vectors;

[0064] The trajectory dwell time vector, attention distribution vector, and operation habit feature vector obtained from the positioning and acquisition module are normalized. Correlation analysis is then performed on the normalized trajectory dwell time vector, operation habit feature vector, and attention distribution vector to generate feature value B. The formula used is as follows:

[0065] ,

[0066] in, , and These represent the normalized trajectory dwell time vector, attention distribution vector, and operation habit feature vector, respectively, with k used to index the elements in the operation habit feature vector. The norm is represented by T, where T is the total inspection time and B is the eigenvalue, used to reflect the comprehensive performance of multi-source behavioral data. , and These eigenvalues ​​reflect the duration of inspection personnel's stays at different locations, their level of attention to different areas of equipment, and the frequency of their operational patterns. In the eigenvalue calculation formula, the exponential function of trajectory dwell time in the numerator enhances the impact of longer dwell times, while the logarithmic function of visual attention mitigates the influence of attention. Operational habits are weighted to the power of 1.5 to suppress abnormal operations, and the denominator adjusts the impact of extreme trajectory dwell times on the overall fusion. If inspection personnel spend a long time in key areas and maintain focused visual attention, the norm values ​​of trajectory and visual attention will increase, leading to an increase in B, representing standardized behavior. Conversely, if there are many abnormal patterns in operational habits, the denominator increases, causing B to decrease, indicating potential anomalies. The eigenvalue B reflects the positive impact of normal behavior and also suppresses the interference of abnormal behavior, improving the accuracy of the analysis.

[0067] By standardizing and unifying the three types of behavioral data—trajectory, line of sight, and operation—and merging them into a comprehensive behavioral feature value B, the overall inspection behavior performance is reflected. A single type of data cannot fully reflect the inspection behavior; fusion can comprehensively consider spatial location, focus, and operation, and comprehensively evaluate the inspection process. The unified behavioral feature value more accurately understands the relationship between inspection behavior and equipment status, improves anomaly detection capabilities, reduces the risk of misjudgment from single data, and enhances the accuracy and robustness of behavioral analysis.

[0068] S5. Collect power plant equipment status parameters and environmental parameters, construct a multivariate nonlinear coupled model, dynamically estimate model parameters through a recursive algorithm, and generate predicted values ​​and residuals of equipment status.

[0069] Collect power plant equipment status parameters and environmental parameters, and generate status parameter vectors. and environmental parameter vector A nonlinear coupled model of Gaussian radial basis functions is constructed to generate predicted equipment conditions. The formula used is as follows:

[0070] ,

[0071] in, This is the weight matrix. The kernel center vector, B is the kernel width, and B is the eigenvalue. Here, M represents the number of Gaussian radial basis function kernels, and m is used to index the Gaussian radial basis function kernels. Based on the recursive least squares algorithm, the model parameters are analyzed to generate predicted equipment status values. and generate residuals The formula used is:

[0072] ,

[0073] in, For the state parameter vector, the predicted device state value The residual is used to reflect the device state parameter vector predicted at time t. This is used to reflect the difference between the actual and predicted equipment conditions. The environmental parameter vector contains environmental information affecting the equipment, such as temperature and humidity. Used to reflect the distance between environmental parameters and the kernel center, the smaller the distance, the greater the weight. The behavior feature adjustment multiplier transforms the intensity and regularity of inspection behavior into its impact on equipment status prediction. Environmental parameters determine the basic trend of equipment status. Equipment performance varies under different environmental conditions. The behavior feature value B reflects the inspection quality. Good behavior makes the prediction more reliable. The kernel function realizes the nonlinear mapping of the environment to the equipment status. At the same time, the behavior feature acts as an amplification adjustment factor to reflect the impact of inspection behavior on equipment maintenance.

[0074] Predicting equipment status based on environmental parameters and comprehensive behavioral characteristics helps determine whether the equipment is functioning properly. Environmental factors, such as temperature and humidity, affect equipment status, while inspection behavior reflects maintenance quality. Combining the two can more accurately predict equipment status. The difference between the predicted value and the actual equipment status can reflect whether the equipment has an anomaly, improving the accuracy of equipment status monitoring. Combining behavioral characteristics reflects the impact of inspection effectiveness on equipment status, which is used for subsequent anomaly judgment and early warning.

[0075] S6. Based on the equipment status residual and the inspection behavior feature vector, perform anomaly judgment, and combine the causal reasoning mechanism to generate anomaly causes and propagation paths based on the equipment and environmental causal graph.

[0076] Construct a causal graph of equipment and environmental variables. Nodes in the graph can be equipment status indicators, environmental parameters, and inspection behavior characteristics. Edges represent the direction and intensity of causal influence between variables. At abnormal nodes, i.e., where abnormal equipment status is detected, trace the causal path of the graph in reverse to identify the antecedent variables that caused the abnormality. Along the directed edges of the causal graph, describe how the abnormality propagates from the cause node to the affected equipment node. The propagation path is used to obtain the abnormality diffusion process.

[0077] Correlation analysis was performed on the residuals and behavioral characteristic values ​​to generate a comprehensive equipment anomaly score. The formula used is:

[0078] ;

[0079] Among them, the comprehensive anomaly score Used to determine the equipment risk index. and The parameters are used to adjust the influence of behavioral characteristics and the contribution of the time integral, respectively, where T is the total inspection time. The 1-norm of the behavioral characteristics represents the degree of behavioral normatization. The more severe the environmental conditions, the greater the error, based on the comprehensive anomaly score. To determine if an anomaly exists, the system generates the cause and propagation path of the anomaly through a cause-effect graph of the device and environment, and outputs a structured anomaly report with a comprehensive anomaly score. The higher the value, the greater the degree of abnormality in the equipment condition and the more irregular the inspection behavior. The smaller the value, the weaker the buffering effect and the larger the anomaly score, reflecting the amplifying effect of non-standard behavior on equipment malfunctions. The integral part also considers environmental parameters; the harsher the environment, the better. The larger the value, the lower the anomaly score will be, reflecting the increased tolerance of the equipment to anomalies in harsh environments; conversely, the anomaly score is more sensitive in better environments.

[0080] Assess the difference between the predicted and actual states of equipment, and combine behavioral characteristics to determine whether there is an anomaly. A large gap between the actual and predicted states indicates that the equipment may be abnormal. At the same time, considering behavioral characteristics can distinguish between anomalies caused by inspection quality and equipment malfunctions, more accurately identify abnormal events, reduce false alarms and missed alarms, and improve safety and equipment reliability.

[0081] This embodiment also provides a computer device applicable to the power plant inspection result analysis method based on artificial intelligence, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the power plant inspection result analysis method based on artificial intelligence as proposed in the above embodiment.

[0082] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0083] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the power plant inspection result analysis method based on artificial intelligence as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0084] In summary, this invention, by constructing an anomaly detection and analysis method based on equipment state residuals and inspection behavior feature vectors, and combining equipment and environmental causal graphs, overcomes the limitations of traditional single-data-dependent anomaly detection. It organically integrates the causal relationship between inspection behavior and equipment state, effectively improving the accuracy and interpretability of anomaly identification, and achieving precise localization of anomaly causes and effective inference of anomaly propagation paths. It should be noted that the above embodiments are only used to illustrate the technical solution of this invention and not to limit it. Although this invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of this invention without departing from the spirit and scope of the technical solution of this invention, and all such modifications and substitutions should be covered within the scope of the claims of this invention.

Claims

1. A method for analyzing power plant inspection results based on artificial intelligence, characterized in that: include, S1. Collect the three-dimensional spatial trajectory data and timestamps of the inspection personnel, and generate a sequence of trajectory points; S2. Based on the collected trajectory point sequence, combined with head posture information, the position of the gaze focus is analyzed to generate the attention distribution matrix of the inspection personnel to the equipment components; S3. Collect the operation event sequence of the inspection personnel, convert it into a symbol sequence based on the time order, extract the operation pattern, and generate an operation habit feature vector; S4. Standardize and fuse the collected three-dimensional spatial trajectory data, timestamps, attention distribution matrix, and operation habit feature vectors to form inspection behavior feature vectors; S5. Collect power plant equipment status parameters and environmental parameters, construct a multivariate nonlinear coupled model, dynamically estimate model parameters through a recursive algorithm, and generate predicted values ​​and residuals of equipment status. S6. Based on the equipment status residual and the inspection behavior feature vector, perform anomaly judgment, and combine the causal reasoning mechanism to generate anomaly causes and propagation paths based on the equipment and environmental causal graph.

2. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 1, characterized in that: In step S1, an ultra-wideband positioning device is used to collect the spatial position of the inspection personnel in real time, obtain three-dimensional coordinate data, sort the collected three-dimensional coordinate data according to the time sequence to form a trajectory point sequence, filter the trajectory point sequence to remove abnormal noise, and obtain smooth trajectory data. The time interval between trajectory data is obtained based on the timestamp to obtain the trajectory dwell time vector, which is used to reflect the trajectory of the inspection space movement state.

3. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 2, characterized in that: In step S2, a head posture sensor is used to collect head posture angle data of the inspection personnel, generating a posture angle sequence. Based on the posture angle sequence and the three-dimensional coordinate data collected by the ultra-wideband positioning device, a gaze direction vector is generated. A fixed gaze projection distance is set, and a correlation analysis is performed on the gaze direction vector and the three-dimensional coordinate data to generate the gaze focus position. The gaze focus position is mapped to a preset equipment area, and the cumulative gaze dwell time in each area is counted. An attention distribution vector is constructed based on the gaze dwell time. The attention distribution vector is used to reflect the degree of attention that the inspection personnel pay to key equipment.

4. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 3, characterized in that: In step S3, the operation events of the inspection personnel are collected. The operation events include operation type and target equipment identifier. An operation event sequence is generated based on the operation events. The operation event sequence is encoded based on time order and converted into a symbol sequence. Correlation analysis is performed on the symbol sequence based on the frequent subsequence mining algorithm to identify high-frequency operation patterns and generate the corresponding occurrence frequency of the operation patterns. An operation habit feature vector is constructed. The operation habit feature vector is used to reflect the operation rules of the inspection personnel.

5. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 4, characterized in that: In step S4, the trajectory dwell time vector, attention distribution vector, and operation habit feature vector obtained by the positioning acquisition module are normalized. Correlation analysis is then performed on the normalized trajectory dwell time vector, operation habit feature vector, and attention distribution vector to generate feature value B. The formula used is as follows: ， in, , and These represent the normalized trajectory dwell time vector, attention distribution vector, and operation habit feature vector, respectively, with k used to index the elements in the operation habit feature vector. The norm is represented by T, the total inspection time is T, and B is the eigenvalue, which is used to reflect the comprehensive performance of multi-source behavioral data.

6. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 5, characterized in that: In step S5, power plant equipment status parameters and environmental parameters are collected, and a status parameter vector is generated. and environmental parameter vector A nonlinear coupled model of Gaussian radial basis functions is constructed to generate predicted equipment conditions. The formula used is as follows: ， in, This is the weight matrix. The kernel center vector, B is the kernel width, and B is the eigenvalue. Here, M represents the number of Gaussian radial basis function kernels, and m is used to index the Gaussian radial basis function kernels. Based on the recursive least squares algorithm, the model parameters are analyzed to generate predicted equipment status values. and generate residuals The formula used is: ， in, For the state parameter vector, the predicted device state value The residual is used to reflect the device state parameter vector predicted at time t. It is used to reflect the difference between the actual state of the equipment and the predicted state.

7. The method for analyzing power plant inspection results based on artificial intelligence as described in claim 6, characterized in that: In step S6, a correlation analysis is performed on the residuals and behavioral characteristic values ​​to generate a comprehensive equipment anomaly score. The formula used is: ， Among them, the comprehensive anomaly score Used to determine the equipment risk index. and The parameters are used to adjust the nonlinear influence of behavioral characteristics on the anomaly score and the contribution of the integral term to the total anomaly score, respectively. T is the total inspection time, based on the comprehensive anomaly score. Determine if an anomaly exists, generate the cause and propagation path of the anomaly through a cause-effect graph of the device and environment, and output a structured anomaly report.

8. A power plant inspection result analysis system based on artificial intelligence, based on the power plant inspection result analysis method based on artificial intelligence as described in any one of claims 1 to 7, characterized in that: include, The positioning and acquisition module is used to collect the three-dimensional spatial trajectory data and timestamps of the inspection personnel and generate a sequence of trajectory points. The gaze capture module is used to analyze the gaze focus position based on the collected trajectory point sequence and head posture information, and generate an attention distribution matrix of the inspection personnel to the equipment components. The operation coding module is used to collect the operation event sequence of inspection personnel, convert it into a symbol sequence based on time order, extract the operation pattern, and generate an operation habit feature vector. The behavior fusion module is used to standardize and fuse the collected three-dimensional spatial trajectory data, timestamps, attention distribution matrix, and operation habit feature vectors to form inspection behavior feature vectors. The state coupling module is used to collect power plant equipment state parameters and environmental parameters, construct a multivariate nonlinear adaptive coupling model, and dynamically estimate the model parameters through a recursive algorithm to obtain the predicted values ​​and residuals of the equipment state. The anomaly analysis module is used to determine anomalies based on equipment status residuals and inspection behavior feature vectors. Combined with a causal reasoning mechanism, it generates anomaly causes and propagation paths based on the causal graph of equipment and environment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the power plant inspection result analysis method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the power plant inspection result analysis method based on artificial intelligence as described in any one of claims 1 to 7.

Images