A power plant inspection robot team emergency management method and system

Abstract

The application relates to the technical field of power plant inspection, in particular to a power plant inspection robot team emergency management method. The method comprises the following steps: collecting different types of data of each device of a power plant during operation; acquiring adjacent data of each data of each type; acquiring the weight of each adjacent data and the characteristic vector of each data, then calculating the characteristic distance between each two data, and acquiring representative data in the data; calculating the average distance of each data; then calculating the deviation degree of the data at the current moment to obtain the abnormal coefficient of the device at present, and then obtaining the inspection priority coefficient of the device; obtaining the fitness degree of a robot and a device based on the shortest route distance of the robot from the device and the inspection priority coefficient of the device; and distributing the devices to be inspected by each robot based on the fitness degree. The application can improve the inspection efficiency of the robot on the power plant.

Description

Technical Field

[0001] This invention relates to the field of power plant inspection technology, specifically to a method and system for emergency management of power plant inspection robots in a team-based manner. Background Technology

[0002] With the continuous expansion of the power industry and the increasing complexity of the operating environment, power plants are placing higher demands on safety, reliability, and response speed in daily operations and emergency response. In recent years, the development of the Internet of Things (IoT) and inspection robot technology has driven a transformation in power plant inspection methods, with inspection robots gradually replacing some manual inspection tasks. Compared to traditional manual methods, inspection robot inspections alleviate, to some extent, problems such as low efficiency, high labor intensity, and slow emergency response.

[0003] Existing inspection robots mostly rely on fixed routes and preset rules to operate, lacking a linkage mechanism with the real-time operating status of equipment. This makes it difficult to dynamically adjust inspection paths and task allocation based on equipment load. For example, when a generator unit experiences abnormal temperature or increased vibration, the inspection robot will still follow its original route, failing to adjust priorities for focused inspections. This can easily lead to missing opportunities to detect potential anomalies, resulting in delayed problem identification. Therefore, how to achieve intelligent task scheduling based on equipment status and other factors to improve inspection efficiency has become a critical issue that power plant inspection and emergency management urgently need to address. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a method and system for team-based emergency management of power plant inspection robots, the specific technical solution of which is as follows:

[0005] In a first aspect, one embodiment of the present invention provides a method for the team-based emergency management of power plant inspection robots, the method comprising:

[0006] Collect different types of data from each piece of equipment in the power plant during operation; obtain neighboring data of a data type based on the acquisition frequency of a data point and the spectrum of that data type.

[0007] The weights of each neighboring data are obtained based on the time interval between a data point and its corresponding neighboring data; the feature vector of a data point is obtained based on its neighboring data, the first-order difference sequence of all neighboring data, and the weights.

[0008] Calculate the feature distance between every two data points based on the feature vectors of each data point in a dataset, and obtain representative data points in that dataset; calculate the mean of the feature distances between a data point and each representative data point in that dataset, and denote it as the average distance of that data point.

[0009] The deviation of the data at the current moment is calculated based on the average distance between the data at the current moment and the average distance between the data at the current moment and the data at the current moment of the representative data; the current anomaly coefficient of a device is obtained based on the deviation of the data at the current moment of various data of a device.

[0010] The inspection priority coefficient of a device is obtained based on its current anomaly coefficient and importance; the compatibility between an inspection robot and a device is obtained based on the shortest route distance between an inspection robot and the device's inspection priority coefficient; and the devices to be inspected by each inspection robot are assigned based on the compatibility.

[0011] Preferably, obtaining neighboring data of a data point based on its acquisition frequency and the spectrum of that data point includes:

[0012] The Fast Fourier Transform is used to process all data in a dataset to obtain the corresponding spectrum. The sampling frequency of the dataset is compared with the frequency corresponding to the maximum amplitude in the spectrum to obtain a first ratio. The difference between the maximum and minimum frequencies in the spectrum is negatively correlated using an exponential function with the natural constant as the base to obtain a first mapping value. The first ratio and the first mapping value are multiplied and rounded up to obtain the number of data extracted from the dataset. Based on the number of data extracted from a data point in the dataset, the neighboring data of that data point are obtained.

[0013] Preferably, the neighboring data of a data point are obtained from the data based on the number of data extracted from that data point, including:

[0014] In a dataset, the number of data extracted from a single data point is determined by sequentially collecting data forward from that data point until the number of collected data points equals the number of data points extracted from that data point. At this point, data collection stops, and the collected data points are considered neighboring data points of that data point.

[0015] Preferably, the weights of each neighboring data are obtained based on the time interval between a data point and its corresponding neighboring data, including:

[0016] The second mapping value of a neighboring data point is obtained by negatively mapping the time interval of a data point to a neighboring data point using an exponential function with the natural constant as the base. The weight of the neighboring data point is obtained by comparing the second mapping value of the neighboring data point with the sum of the second mapping values ​​of all the neighboring data points of the data point.

[0017] Preferably, the feature vector of a data point is obtained based on its neighboring data and the first-order difference sequence and weights corresponding to all neighboring data points, including:

[0018] The first-order difference mean of a data point is obtained by averaging the first-order difference values ​​in the first-order difference sequence corresponding to all its neighboring data points, using the weights of the neighboring data points excluding the first neighboring data point. The weighted average of these neighboring data points is then calculated using the weights of the neighboring data points themselves, yielding the weighted average. The weighted standard deviation of each neighboring data point is calculated using its weights, and the coefficient of variation (COP) is calculated using the COP and the COP. The first-order difference mean, the COP, and the COP constitute the feature vector of the data.

[0019] Preferably, obtaining representative data from this type of data includes:

[0020] A representative index for a given dataset is obtained by negatively mapping the mean characteristic distance between a data point and other data points using an exponential function with the natural constant as the base. The data point with the largest representative index is then selected as the representative data.

[0021] Preferably, the deviation of the data at the current moment is calculated based on the average distance between the data at the current moment and the average distance between the data at the current moment and the data at the current moment, including:

[0022] The average distances of the current data's neighboring data are arranged in a time sequence and denoted as the average distance sequence of the current data. The average slope between every two average distances in the average distance sequence of the current data is weighted and averaged using the weights of the neighboring data (excluding the first neighbor). Similarly, the weighted slopes of representative data are obtained. A third mapping value is obtained by mapping the average difference between the weighted slope of the current data and the weighted slopes of representative data using an exponential function with a base of the natural constant. The deviation of the current data from the average distance of the current data is then multiplied by the third mapping value to obtain the deviation of the current data.

[0023] Preferably, the current anomaly coefficient of a device is obtained based on the degree of deviation of various data at the current moment, including:

[0024] The deviation of the various data corresponding to the device at the current moment is normalized to form a dataset, which is denoted as the deviation dataset. The average deviation is obtained by averaging the elements of the deviation dataset using the weights of various data. The current anomaly coefficient of the device is obtained by multiplying the average deviation, the maximum value in the deviation dataset, and the weight of the data corresponding to the maximum value.

[0025] Preferably, the inspection priority coefficient of a device is obtained based on its current anomaly coefficient and importance, including:

[0026] The inspection priority coefficient of a device is obtained by multiplying its current anomaly coefficient by its importance coefficient.

[0027] Preferably, the compatibility between the inspection robot and the device is determined based on the shortest route distance between the inspection robot and the device and the inspection priority coefficient of the device, including:

[0028] Normalize the shortest path distance between each inspection robot and each device, as well as the inspection priority coefficient of each device, to obtain the normalized shortest path distance and the normalized inspection priority coefficient. Subtract the normalized shortest path distance between an inspection robot and a device from the first preset value to obtain the distance feature coefficient. Perform a weighted sum of the distance feature coefficient and the normalized inspection priority coefficient of the device to obtain the fit between the inspection robot and the device.

[0029] Preferably, the equipment to be inspected by each inspection robot is assigned based on the adaptability, including:

[0030] An adaptation matrix is ​​constructed using the current adaptation degree between each inspection robot and each device. The Hungarian algorithm is then used to solve the adaptation matrix to obtain the optimal matching relationship between the inspection robot and the device, thus identifying the devices that each inspection robot needs to inspect.

[0031] Secondly, the present invention also provides a power plant inspection robot team-based emergency management system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of a power plant inspection robot team-based emergency management method.

[0032] The embodiments of the present invention have at least the following beneficial effects: This application collects different types of data from each piece of equipment in a power plant during operation, and then obtains neighboring data for each piece of data; then obtains the feature vector of each piece of data, and calculates the feature distance between every two pieces of data of the same type based on the feature vector, and obtains representative data for each type of data; calculates the average feature distance between a piece of data and each representative piece of data of that type, and records it as the average distance of that data; then analyzes the representative data for each type of data to obtain the deviation degree of each piece of data at the current moment; then obtains the current anomaly coefficient of the equipment based on the deviation degree of various data of a piece of equipment at the current moment; then, combined with the importance of the equipment, obtains the inspection priority coefficient of the equipment; finally, combined with the inspection priority coefficient of the equipment and the shortest route distance between the inspection robot and the equipment, obtains the fit degree between the inspection robot and the equipment, and then allocates the equipment to be inspected by each inspection robot based on the fit degree, taking into account both the distance between the inspection robot and the equipment and the inspection priority of the equipment, making the inspection matching relationship between the inspection robot and the equipment more reasonable, improving the working efficiency of the inspection robot and the timeliness and efficiency of equipment maintenance, etc. Attached Figure Description

[0033] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0034] Figure 1 A flowchart illustrating a method for emergency management of power plant inspection robots in a team-based manner, as provided in an embodiment of the present invention. Detailed Implementation

[0035] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a power plant inspection robot-based team-based emergency management method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0037] The following description, in conjunction with the accompanying drawings, details a specific scheme for a team-based emergency management method for power plant inspection robots provided by this invention.

[0038] Example 1: The main application scenarios of this invention are:

[0039] This application aims to improve the intelligent scheduling capabilities of inspection robots in the working environment of power plants, thereby increasing the efficiency of inspection robots.

[0040] Please see Figure 1 The diagram illustrates a method flowchart for a power plant inspection robot-based emergency management method according to an embodiment of the present invention. The method includes the following steps:

[0041] Step S1: Collect different types of data from each piece of equipment in the power plant during operation; obtain neighboring data of a data type based on the acquisition frequency of a data type and the spectrum of that data type.

[0042] When inspection robots conduct inspections at power plants, current scheduling methods still struggle to achieve real-time integration of robot and human operation statuses. They lack digital modeling and intelligent management of team-based organizational structures, hindering efficient and orderly emergency response. Therefore, there is an urgent need to build a human-like team-based management mechanism, utilizing intelligent methods to coordinate the inspection and emergency operations of inspection robots and human staff, in order to meet the needs of power plant digital transformation and safe production.

[0043] This application primarily aims to achieve real-time data collection during the operation of equipment in power plants, enabling analysis of their working status and subsequent task scheduling and allocation for inspection robots. Therefore, it is necessary to collect various types of data from the operating equipment within the power plant, such as the operating temperature, vibration amplitude, current, and voltage of generator sets, and the water flow rate of the cooling system. Different types of data correspond to different equipment operation conditions, and the specific types of data collected can be flexibly selected based on the needs of the monitoring equipment and the site conditions. The collected data is transmitted to the task scheduling and control terminal of the inspection robot via wired or wireless means, thus providing reliable data support for subsequent task scheduling and path optimization.

[0044] Power plants typically contain numerous and diverse pieces of equipment. To ensure their continuous and stable operation, inspection robots need to periodically check these devices. However, the operating status of equipment often varies at different times. For example, units operate at high loads during the day and may enter a low-load or standby state at night. If inspection robots always perform inspections at fixed intervals and along predetermined routes, it not only easily leads to unnecessary repetitive work but may also overlook potential anomalies in certain equipment under specific conditions. Taking the generator cooling system as an example, temperature anomalies are more likely to occur when the load increases. If the inspection frequency and path are not dynamically adjusted, the optimal time for anomaly detection can be easily delayed. Therefore, the traditional timed and fixed inspection mode suffers from low efficiency, resource waste, and delayed response. Thus, it is necessary to combine the real-time operating conditions of the equipment to intelligently plan and schedule inspection tasks, thereby improving inspection efficiency while prioritizing the monitoring of key risk points, further enhancing the safety and resource utilization of power plant operations.

[0045] Therefore, firstly, for this scenario, considering the changing characteristics of different data collected by different devices, the range for analysis needs to be dynamically determined based on the overall frequency of change. For example, a shorter analysis range is used for rapidly fluctuating data to capture the sensitivity of local changes, while a longer range is used for slowly changing data to obtain more reference data and improve the stability of the analysis, thereby achieving feature analysis and detection that better fits the working conditions.

[0046] For a single data point from different types of data from a single device, it is necessary to determine its analysis range and then obtain its neighboring data. This is done by using the sampling frequency of a data point within a given data type and its spectrogram to obtain its neighboring data.

[0047] Specifically, the Fast Fourier Transform is used to process all data in a dataset to obtain the corresponding spectrum; the sampling frequency of the dataset is compared with the frequency corresponding to the maximum amplitude in the spectrum to obtain a first ratio; the difference between the maximum and minimum frequencies in the spectrum is negatively correlated using an exponential function with the natural constant as the base to obtain a first mapping value; the first ratio and the first mapping value are multiplied and rounded up to obtain the number of data extracted from the dataset.

[0048] The neighboring data of a given data point is determined by the number of data extracted from it. Specifically, within a given dataset, starting with the data point extracted from that data point and proceeding backwards, data is collected sequentially until the number of collected data points equals the number of data extracted from that data point. This collected data is then considered the neighboring data of that data point. For example, if there are four data points, and the fourth data point has a data extraction count of 3, then the neighboring data of the fourth data point are the second, third, and fourth data points. The neighboring data of a given data point includes the data point itself.

[0049] The specific calculation model for the number of data extracted is as follows:

[0050] ,

[0051] Where W represents the number of data points extracted from a data type of a device, and f represents the collection frequency when collecting this data; This indicates the frequency corresponding to the maximum amplitude in the spectrum of a certain type of data, representing the main frequency variation of the signal; ⌈ ⌉ indicates the rounding up sign; Represents an exponential function with the natural constant as its base; the first ratio. This indicates the number of data points corresponding to the main frequency of change in the collected data (i.e., the number of data points contained within a main period). If the frequency variations in this type of data are more abundant, then the number of data points extracted for other data analysis should be smaller. Otherwise, the selected data range should be closer to the main frequency of change, i.e., it only needs to include data within one period of change. RS represents the frequency range of the spectrum of this type of data (the difference between the maximum and minimum frequencies). The larger this value, the more abundant the frequencies in the data, and therefore the smaller the selected data analysis range. Thus, the larger the first mapping value exp(-RS), the larger the number of data points extracted for each data point during analysis, and the more neighboring data obtained; conversely, the smaller the value, the less abundant the frequency variations.

[0052] Step S2: Obtain the weights of each neighboring data based on the time interval between a data point and its corresponding neighboring data points; obtain the feature vector of a data point based on its neighboring data points and the first-order difference sequence and weights corresponding to all neighboring data points.

[0053] The above steps obtain the neighboring data corresponding to each data point. Further, to highlight the local characteristics of each data point within a dataset, the weights of each neighboring data point are calculated when analyzing a given data point. A second mapping value for the neighboring data point is obtained by negatively mapping the time interval between the data point and its neighboring data using an exponential function with a base of the natural constant. The weight of the neighboring data point is then obtained by comparing its second mapping value with the sum of the second mapping values ​​of all its neighboring data points.

[0054] The specific model for calculating the weights is as follows:

[0055]

[0056] in, This represents the weight of the s-th neighboring data among the neighboring data of the i-th data in a dataset. This represents the time interval between the s-th neighboring data and the i-th data; This represents the second mapping value of the s-th neighboring data; the smaller the time interval, the greater the corresponding weight. This represents the number of neighboring data of the i-th data in a dataset.

[0057] Furthermore, a feature vector of a data point is obtained based on its neighboring data and the first-order difference sequence and weights corresponding to all neighboring data points.

[0058] Specifically, the neighboring data of a given data point are arranged in chronological order, and their first-order differences are obtained to form a first-order difference sequence corresponding to all neighboring data points. The first-order difference values ​​in the first-order difference sequence corresponding to all neighboring data points are weighted and averaged using the weights of the neighboring data points excluding the first neighboring data point, yielding the first-order difference mean of the given data point. The weighted average of the neighboring data points is then calculated using the weights of the neighboring data points, yielding the weighted data mean of the given data point. The weighted standard deviation of each neighboring data point is calculated using the weighted standard deviation and the weighted data mean, and the coefficient of variation is calculated using the weighted standard deviation and the weighted data mean, denoted as the weighted coefficient of variation of the given data point. The first-order difference mean, the data mean, and the weighted coefficient of variation of the given data point constitute the feature vector of the given data point.

[0059] The coefficient of variation is obtained using existing methods, and the formula for calculating the weighted standard deviation is:

[0060] ,

[0061] Where σ is the weighted standard deviation; This represents the number of neighboring data of the i-th data in a dataset; This represents the weight of the s-th neighboring data among the neighboring data of the i-th data in a dataset. Let μ represent the s-th neighboring data among the neighboring data of the i-th data in a dataset, and let μ represent the mean of the neighboring data of the i-th data.

[0062] This allows us to obtain the feature vector for each data point in each data set.

[0063] Step S3: Calculate the feature distance between every two data points based on the feature vectors of each data point in a dataset, and obtain representative data in that dataset; calculate the mean of the feature distances between a data point and each representative data point in that dataset, and denot it as the average distance of that data point.

[0064] The above steps obtain the feature vector of each data point. Further, based on the Euclidean distance calculation method, the feature distance between any two data points is calculated using their feature vectors. For any given set of data, the representative data points are identified, which likely represent the characteristic state of the data at most times. Then, a representative value from the currently collected data is selected for subsequent reference, thereby identifying potential anomalies in the data and facilitating decision-making for the inspection robot.

[0065] Specifically, a representative index for a given dataset is obtained by negatively mapping the mean of the characteristic distances between a data point and other data points using an exponential function with the natural constant as the base. The data point with the largest representative index is then selected as the representative data.

[0066] The specific calculation model for the representative indicator is as follows:

[0067] ,

[0068] in, This represents a representative indicator of the a-th data point in a dataset. This represents the feature distance between the a-th and r-th data points in this type of data; n represents the number of other data points in this type of data besides the a-th data point. The smaller the value, the smaller the feature distance between the a-th data point and other data points, and the greater its representation of the features of the entire data. The larger the value, the more representative the a-th data point is.

[0069] This allows us to obtain representative data for each type of data, where there may be multiple representative data. We then calculate the mean of the feature distances between a given data point and all its representative data points, denoted as the average distance of that data point.

[0070] Step S4: Calculate the deviation of the data at the current moment based on the average distance between the data at the current moment and the average distance between the data at the current moment and the data at the current moment of the representative data; obtain the current anomaly coefficient of the device based on the deviation of the data at the current moment of various data of the device.

[0071] The above steps obtain the difference (average distance) between each data point and the representative data. Further analysis of the current data point reveals its degree of deviation from the representative data. The degree of deviation of the current data point is calculated based on the average distances between the current data point's neighboring data and the average distances between the neighboring data points of each representative data point.

[0072] Specifically, the average distances of the data at the current moment are arranged into a sequence according to time, denoted as the average distance sequence of the data at the current moment. The average slope between every two average distances in the average distance sequence of the data at the current moment is weighted and averaged using the weights of the other neighboring data (excluding the first neighboring data). Similarly, the weighted slopes corresponding to representative data of this type are obtained. A third mapping value is obtained by mapping the average difference between the weighted slope of the data at the current moment and the weighted slopes corresponding to representative data of this type using an exponential function with a base of the natural constant. The deviation of the data at the current moment is obtained by multiplying the third mapping value by the average distance of the data at the current moment.

[0073] Deviation refers to the degree of deviation between the current data and representative data of that type of data. The specific calculation model for deviation is as follows:

[0074] ,

[0075] Where Y represents the degree of deviation of the data at the current moment; e represents the natural constant; k represents the weighted slope corresponding to the data at the current moment. For example, if a data has 4 neighboring data, then there are 4 average distances in the corresponding average distance sequence. At this time, there are three slopes. Then, the average slope is obtained by weighting the three slopes using the weights of the last three neighboring data. This represents the weighted slope corresponding to the j-th representative data point in this type of data, and g represents the number of representative data points in this type of data. This is the third mapping value. The larger this value is, the greater the potential abnormal trend between the pattern change at this moment and the pattern at the reference moment.

[0076] This allows us to obtain the degree of deviation of each type of data at the current moment. Furthermore, for a device, different types of data may reflect different abnormal states of the device and have different monitoring importance. Therefore, different weights can be set for various data in the device (the value is between 0 and 1, which represents the ability to show abnormal state of the device, and can be determined by the operation and maintenance personnel. For example, the temperature data in the device is more important and can better highlight the state of the device, so its weight is higher. Otherwise, a lower weight can be set. The sum of the weights of these data is 1. By default, the weights of each type of data are equal).

[0077] Furthermore, the current anomaly coefficient of a device is obtained based on the deviation of its various data at the current moment. Specifically, the deviation of the various data corresponding to the device at the current moment is normalized and formed into a dataset, denoted as the deviation dataset; the average deviation is obtained by averaging the elements of the deviation dataset using the weights of various data; the current anomaly coefficient of the device is obtained by multiplying the average deviation, the maximum value in the deviation dataset, and the weight of the data corresponding to the maximum value.

[0078] The specific calculation model for the anomaly coefficient is as follows:

[0079] ,

[0080] Where W represents the anomaly coefficient of a device; This represents a dataset composed of normalized deviations from various data points corresponding to the device at the current moment; it is also known as the deviation dataset. This represents the maximum value in the dataset indicating the degree of deviation. This represents the weight of a data point corresponding to the maximum value in the dataset, indicating the degree of deviation. For different types of data, the deviation of the data at this time is the maximum value of the deviation in the dataset, multiplied by the weight of that type of data. The larger this value is, the more abnormal the data of that type of data is. This represents the degree of deviation corresponding to the r-th data type in the deviation dataset. This represents the weight corresponding to the r-th type of data. This represents the average degree of deviation; the larger the value, the more abnormal the device as a whole may be.

[0081] This allows us to obtain the current anomaly coefficient for each device.

[0082] Step S5: Obtain the inspection priority coefficient of a device based on its current anomaly coefficient and importance; obtain the compatibility between an inspection robot and a device based on the shortest route distance between an inspection robot and the device's inspection priority coefficient; and assign the devices to be inspected by each inspection robot based on the compatibility.

[0083] The above describes obtaining the current anomaly coefficient for each piece of equipment. This coefficient, combined with the equipment's importance, allows us to determine its inspection priority coefficient. Specifically, the inspection priority coefficient is obtained by multiplying the current anomaly coefficient of a piece of equipment by its importance. A higher anomaly coefficient indicates greater importance and a higher inspection priority. The importance of each piece of equipment is pre-set manually, with values ​​ranging from 0 to 1. For example, core equipment such as main transformers, generators, and boilers are directly related to the safe and stable operation of the power plant. Failures in these systems could lead to a complete plant shutdown or significant economic losses; therefore, their maintenance priority must be significantly higher than that of general auxiliary equipment, such as lighting systems or cooling fans. By pre-setting different importance levels for different equipment, the inspection robot can prioritize the inspection and maintenance of high-risk, high-value equipment during task scheduling, thereby improving the overall reliability and safety of the power plant's operation.

[0084] Furthermore, in the power plant inspection scenario described in this application, the number of devices is large, their distribution is wide, and their operating states are complex. If a fixed area division mode is adopted, it is easy for some inspection robots to be overloaded while others are idle, resulting in reduced inspection efficiency. Therefore, this application adopts a global parallel inspection mode, where all inspection robots can freely inspect throughout the entire power plant without area restrictions. All inspection robots can inspect throughout the entire power plant, and the system uses a fitness matrix and the Hungarian algorithm for global task allocation. The inspection robot with the highest fitness with the device is prioritized for allocation. This method achieves optimal overall allocation, avoiding resource imbalances in the area division mode, and is particularly suitable for power plants with many sudden tasks and significant differences in device priority.

[0085] This allows us to obtain the shortest path distance between each inspection robot and each device at the current moment. The shortest path distance refers to the shortest distance that an inspection robot travels to a device. The shortest path distances between each inspection robot and each device, as well as the inspection priority coefficients of each device, are then normalized to obtain normalized shortest path distances and normalized inspection priority coefficients. This eliminates the difference in dimensions between these two sets of data.

[0086] Furthermore, by subtracting the normalized shortest path distance between an inspection robot and a device from the first preset value, a distance feature coefficient is obtained; the distance feature coefficient and the normalized inspection priority coefficient of the device are weighted and summed to obtain the compatibility between the inspection robot and the device.

[0087] The specific calculation model for fitness is as follows:

[0088] ,

[0089] in, This represents the compatibility between the current inspection robot a and the equipment b. β represents the parameter weight, used to balance the relative influence of "inspection priority coefficient" and "shortest route distance" in the compatibility calculation. Its value range is recommended to be 0.5 to 1, with a default value of 0.7. This is because in power plant operation scenarios, the operating status of the equipment and the inspection priority are often more important than the movement cost of the inspection robot. Therefore, it needs to be given a higher weight in the calculation. However, the specific value of β can be flexibly adjusted according to actual needs to adapt to the differences in the focus of task allocation in different scenarios. norm(D(a,b)) represents the normalized shortest route distance between inspection robot a and equipment b. 1 is the first preset value, and 1-norm(D(a,b)) is the distance feature coefficient. The larger this value is, the higher the compatibility between the inspection robot and the equipment to be maintained in establishing task links. This represents the inspection priority coefficient for device b. This represents the normalized inspection priority coefficient. The larger the value, the greater the compatibility with the inspection robot.

[0090] This allows us to obtain the fit degree between each inspection robot and each piece of equipment, and then construct a fit degree matrix of size A×B, where A represents the number of inspection robots and B represents the number of maintenance devices. Subsequently, this fit degree matrix is ​​transformed into a matching problem, which is solved using the Hungarian algorithm to obtain the optimal matching relationship between robots and equipment. This allows us to determine the equipment each inspection robot needs to inspect and assign tasks accordingly. It should be noted that the Hungarian algorithm, as a mature and widely used optimal matching method, has principles and steps that are well-known and will not be elaborated upon in this paper.

[0091] Example 2: This example provides a power plant inspection robot team-based emergency management system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of a power plant inspection robot team-based emergency management method. Since Example 1 has already described a power plant inspection robot team-based emergency management method in detail, it will not be elaborated upon here.

[0092] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0093] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0094] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for team-based emergency management of power plant inspection robots, characterized in that, The method includes: Collect different types of data from each piece of equipment in the power plant during operation; obtain neighboring data of a data type based on the acquisition frequency of a data point and the spectrum of that data type. The weights of each neighboring data are obtained based on the time interval between a data point and its corresponding neighboring data; the feature vector of a data point is obtained based on its neighboring data, the first-order difference sequence of all neighboring data, and the weights. Calculate the feature distance between every two data points based on the feature vectors of each data point in a dataset, and obtain representative data points in that dataset; calculate the mean of the feature distances between a data point and each representative data point in that dataset, and denote it as the average distance of that data point. The deviation of the data at the current moment is calculated based on the average distance between the data at the current moment and the average distance between the data at the current moment and the data at the current moment of the representative data; the current anomaly coefficient of a device is obtained based on the deviation of the data at the current moment of various data of a device. The inspection priority coefficient of a device is obtained based on its current anomaly coefficient and importance; the compatibility between an inspection robot and a device is obtained based on the shortest route distance between an inspection robot and the device's inspection priority coefficient; and the devices to be inspected by each inspection robot are assigned based on the compatibility. The calculation of the deviation of the data at the current moment based on the average distance between the data at the current moment and the average distance between the data at the current moment and the data at the current moment includes: The average distances of the current data's neighboring data are arranged in a time sequence and denoted as the average distance sequence of the current data. The slopes between any two average distances in the average distance sequence of the current data are weighted and averaged using the weights of the neighboring data (excluding the first neighbor). Similarly, the weighted slopes of representative data are obtained. A third mapping value is obtained by mapping the average difference between the weighted slope of the current data and the weighted slopes of representative data using an exponential function with a base of the natural constant. The deviation of the current data from the average distance of the current data is then calculated by multiplying the third mapping value by the average distance of the current data. The allocation of equipment to be inspected by each inspection robot based on the adaptability includes: An adaptation matrix is ​​constructed using the current adaptation degree between each inspection robot and each device. The Hungarian algorithm is then used to solve the adaptation matrix to obtain the optimal matching relationship between the inspection robot and the device, thus identifying the devices that each inspection robot needs to inspect.

2. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The step of obtaining neighboring data of a data type based on the acquisition frequency of that data type and the spectrum of that data type includes: The Fast Fourier Transform is used to process all data in a dataset to obtain the corresponding spectrum. The sampling frequency of the dataset is compared with the frequency corresponding to the maximum amplitude in the spectrum to obtain a first ratio. The difference between the maximum and minimum frequencies in the spectrum is negatively correlated using an exponential function with the natural constant as the base to obtain a first mapping value. The first ratio and the first mapping value are multiplied and rounded up to obtain the number of data extracted from the dataset. Based on the number of data extracted from a data point in the dataset, the neighboring data of that data point are obtained.

3. The emergency management method for power plant inspection robots based on teamwork as described in claim 2, characterized in that, The step of obtaining neighboring data of a data point in the data type based on the number of data extracted from that data type includes: In a dataset, the number of data extracted from a single data point is determined by sequentially collecting data forward from that data point until the number of collected data points equals the number of data points extracted from that data point. At this point, data collection stops, and the collected data points are considered neighboring data points of that data point.

4. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The step of obtaining the weights of each neighboring data based on the time interval between a data point and its corresponding neighboring data includes: The second mapping value of a neighboring data point is obtained by negatively mapping the time interval of a data point to a neighboring data point using an exponential function with the natural constant as the base. The weight of the neighboring data point is obtained by comparing the second mapping value of the neighboring data point with the sum of the second mapping values ​​of all the neighboring data points of the data point.

5. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The process of obtaining a feature vector for a data point based on its neighboring data, the first-order difference sequence corresponding to all neighboring data, and weights includes: The first-order difference mean of a data point is obtained by averaging the first-order difference values ​​in the first-order difference sequence corresponding to all its neighboring data points, using the weights of the neighboring data points excluding the first neighboring data point. The weighted average of these neighboring data points is then calculated using the weights of the neighboring data points themselves, yielding the weighted average. The weighted standard deviation of each neighboring data point is calculated using its weights, and the coefficient of variation (COP) is calculated using the COP and the COP. The first-order difference mean, the COP, and the COP constitute the feature vector of the data.

6. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The acquisition of representative data from this type of data includes: A representative index for a given dataset is obtained by negatively mapping the mean characteristic distance between a data point and other data points using an exponential function with the natural constant as the base. The data point with the largest representative index is then selected as the representative data.

7. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The process of obtaining the current anomaly coefficient of a device based on the deviation of various data at the current moment includes: The deviation of the various data corresponding to the device at the current moment is normalized to form a dataset, which is denoted as the deviation dataset. The average deviation is obtained by averaging the elements of the deviation dataset using the weights of various data. The current anomaly coefficient of the device is obtained by multiplying the average deviation, the maximum value in the deviation dataset, and the weight of the data corresponding to the maximum value.

8. The emergency management method for power plant inspection robots based on teamwork as described in claim 1, characterized in that, The process of obtaining the inspection priority coefficient of a device based on its current anomaly coefficient and importance includes: The inspection priority coefficient of a device is obtained by multiplying its current anomaly coefficient by its importance coefficient.

9. A method for team-based emergency management of power plant inspection robots according to claim 1, characterized in that, The degree of compatibility between an inspection robot and a device is determined based on the shortest route distance between the robot and the device and the device's inspection priority coefficient. This includes: Normalize the shortest path distance between each inspection robot and each device, as well as the inspection priority coefficient of each device, to obtain the normalized shortest path distance and the normalized inspection priority coefficient. Subtract the normalized shortest path distance between an inspection robot and a device from the first preset value to obtain the distance feature coefficient. Perform a weighted sum of the distance feature coefficient and the normalized inspection priority coefficient of the device to obtain the fit between the inspection robot and the device.

10. A power plant inspection robot team-based emergency management system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the emergency management method for a power plant inspection robot team as described in any one of claims 1-9.

Images