Method and System for Operational Rounding Inspection and Trend Analysis of Power Generation Plant Equipment

Abstract

The present application relates to the technical field of data processing, and provides a method and a system for the operation patrol inspection and trend analysis of power plant facilities. In the process of the patrol inspection camera's patrol, the video capture device is activated to obtain the real-time video information of the power plant facilities, which is input into the image processing recognition model to obtain the equipment status recognition information. Connected to the target power plant facilities, upload data is obtained for equipment status prediction, and the equipment status prediction result is obtained. A patrol inspection task is set, and the patrol inspection camera is made to perform an automatic patrol inspection on the target power plant facilities according to the patrol inspection task, thereby solving the technical problems that the efficiency of the operation patrol inspection of the power plant facilities is low and the accuracy is not high, and the reliability of the operation patrol inspection information of the power plant facilities is low. A recognition model is constructed to improve the accuracy, the patrol inspection route in the patrol inspection task of the patrol inspection camera is optimized by using the equipment status prediction, and an automatic patrol inspection is performed according to the patrol inspection task, so that the technical effect of improving the operation patrol inspection efficiency and maintaining the reliability of the operation patrol inspection information can be realized.

Description

Technical Field

[0001] This application relates to the technical field related to data processing, and specifically relates to a method and system for the operation patrol inspection and trend analysis of power plant equipment.

Background Art

[0002] In recent years, with the large-scale popularization of unmanned power plants in the power grid, smart patrol cameras have gradually replaced manual labor to conduct patrol inspections of power plant equipment. Most of the instrument equipment to be patrolled is pointer-type instrument equipment, such as air pressure gauges, thermometers, oil temperature gauges, lightning arrester instruments, etc. This type of instrument has a simple structure, low manufacturing cost, convenient maintenance, strong electromagnetic interference resistance, high reliability, and has the characteristics of dust-proof, waterproof, and anti-freezing, and is widely applied in the power system.

[0003] Since most of the instrument equipment is pointer-type instrument equipment, the output result is not a digital signal and cannot be directly input into the computer system. The recognition of the conventional instrument readings is mostly by manual discrimination. The recognition by human eyes is easily affected by many human factors, with low detection efficiency, low accuracy, and there are also cases where manual interpretation cannot be performed, such as in cases with high radiation or high danger.

[0004] From the above, there are technical problems in the prior art that the operation patrol inspection efficiency of power plant equipment is low and the accuracy is not high, and the reliability of the information on the operation patrol inspection of power plant equipment is low.

Summary of the Invention

Problems to be Solved by the Invention

[0005] In order to solve the technical problems in the prior art that the operation patrol inspection efficiency of power plant equipment is low and the accuracy is not high, and the reliability of the information on the operation patrol inspection of power plant equipment is low, this application provides a method and system for the operation patrol inspection and trend analysis of power plant equipment.

Means for Solving the Problems

[0006] In view of the above problems, embodiments of the present application provide a method and system for the operation patrol inspection and trend analysis of power plant equipment.

[0007] One aspect of the present disclosure provides a method for the operation patrol inspection and trend analysis of power plant equipment. The method is applicable to a system for operation patrol inspection and trend analysis. The system for operation patrol inspection and trend analysis and the patrol inspection camera are communicatively interconnected. The method includes, during the process of the patrol inspection camera's patrol, activating the video capture equipment to obtain real-time video information of the power plant equipment; inputting the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment state recognition information; connecting to the target power plant equipment to obtain upload data; performing equipment state prediction based on the upload data according to the equipment state recognition information to obtain an equipment state prediction result; setting a patrol inspection task according to the equipment state prediction result; and causing the patrol inspection camera to perform an automatic patrol inspection on the target power plant equipment according to the patrol inspection task.

[0008] Another aspect of the present disclosure provides a system for the operation patrol inspection and trend analysis of power plant equipment. The system includes a video information acquisition module for activating video capture equipment to obtain real-time video information of power plant equipment during the process of the patrol inspection camera's patrol; an equipment state recognition module for inputting the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment state recognition information; an upload data acquisition module for connecting to the target power plant equipment to obtain upload data; an equipment state prediction module for performing equipment state prediction based on the upload data according to the equipment state recognition information to obtain an equipment state prediction result; a patrol inspection task setting module for setting a patrol inspection task according to the equipment state prediction result; and an automatic patrol inspection module for causing the patrol inspection camera to perform an automatic patrol inspection on the target power plant equipment according to the patrol inspection task.

Advantages of the Invention

[0009] One or more technical means provided in the present application have at least the following technical effects or advantages.

[0010] In the process of the patrol camera's patrol, the present application activates the video capture equipment, obtains the real-time video information of the power plant equipment, inputs the real-time video information of the power plant equipment into the image processing recognition model, obtains the equipment status recognition information, connects to the target power plant equipment, obtains the upload data, performs equipment status prediction based on the upload data according to the equipment status recognition information, obtains the equipment status prediction result, sets the patrol inspection task, and causes the patrol camera to perform automatic patrol inspection on the target power plant equipment according to the patrol inspection task, thereby constructing a recognition model and improving the accuracy, optimizing the patrol route in the patrol inspection task of the patrol camera by using the equipment status prediction, performing automatic patrol inspection according to the patrol inspection task, and realizing the technical effect of improving the operation patrol inspection efficiency and maintaining the reliability of the operation patrol inspection information.

[0011] The above description is only an overview of the technical means of the present application. To more clearly understand the technical means of the present application, implement it according to the content of the specification, and more clearly understand the above and other objects, features, and advantages of the present application, the specific embodiments of the present application are listed below.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Modes for Carrying Out the Invention

[0013] The embodiments of the present application provide a method and a system for the operation patrol inspection and trend analysis of power plant equipment, thereby solving the technical problems that the efficiency of the operation patrol inspection of power plant equipment is low and the accuracy is not high, and the reliability of the operation patrol inspection information of power plant equipment is low. A recognition model is constructed to improve the accuracy, and the patrol inspection route in the patrol inspection tasks of the patrol inspection camera is optimized by using equipment status prediction, and automatic patrol inspection is performed according to the patrol inspection tasks, so that the technical effect of improving the operation patrol inspection efficiency and maintaining the reliability of the operation patrol inspection information can be realized.

[0014] After introducing the basic principle of the present application, the following will specifically describe various non-limiting embodiments of the present application in conjunction with the accompanying drawings.

[0015] Embodiment 1

[0016] As shown in FIG. 1, the embodiment of the present application provides a method for the operation patrol inspection and trend analysis of power plant equipment. The method is applicable to a system for operation patrol inspection and trend analysis. The system for operation patrol inspection and trend analysis and the patrol inspection camera are communicatively interconnected. The method includes the following.

[0017] S10: In the process of the patrol inspection camera patrolling, start the video capture equipment to obtain the real-time video information of the power plant equipment.

[0018] S20: Input the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment status recognition information.

[0019] Step S20 includes the following steps.

[0020] S21: Extract a plurality of sample image information synchronously at a previous time when the target power plant equipment is in different equipment states.

[0021] S22: Identify corresponding multiple sample equipment status recognition information based on the multiple sample image information.

[0022] S23: Employ the plurality of sample image information and the plurality of sample equipment state recognition information to construct the image processing recognition model.

[0023] S24: Input the real-time video information of the power plant equipment into the image processing recognition model to obtain the equipment state recognition information.

[0024] The operation patrol inspection and trend analysis system and the patrol inspection camera are communicatively interconnected. Specifically, the communicative interconnection is, simply speaking, to communicate through signal transmission, and a communication network is configured between the operation patrol inspection and trend analysis system and the patrol inspection camera. During the process of the patrol inspection camera's patrol (various sensors such as a visible light sensor, an infrared sensor, and an audio sensor are attached to the patrol inspection camera), the video capture equipment is activated, and the video capture equipment is used to capture the audio and image information of the target power plant equipment, obtain the real-time video information of the power plant equipment, and provide data support for subsequent processing.

[0025] Inputting the real-time video information of the power plant equipment into the image processing recognition model to obtain the equipment state recognition information specifically includes: based on the operation patrol inspection and trend analysis system, when the target power plant equipment was in different equipment states at a previous time, synchronously extracting a plurality of sample image information (the plurality of sample image information corresponds one-to-one to different equipment states); based on the plurality of sample image information, specifying a corresponding plurality of sample equipment state recognition information from the operation log of the target power plant equipment; adopting the plurality of sample image information and the plurality of sample equipment state recognition information to construct the image processing recognition model; inputting the real-time video information of the power plant equipment into the image processing recognition model as input data, and outputting the equipment state recognition information (the equipment state recognition information includes notation indication information and the display state of the enclosure), providing a reference for subsequent substitution into the image processing recognition model.

[0026] Step S23 includes the following steps.

[0027] S231: Based on the convolutional neural network as the basis of the model, construct the image processing recognition model. The input data of the image processing recognition model is the real-time video information of the power plant equipment, and the output data is the equipment status recognition information.

[0028] S232: Adopt the plurality of sample image information and the plurality of sample equipment status recognition information, decompose them to obtain a training information set and a verification information set.

[0029] S233: Set the first model accuracy threshold.

[0030] S234: Use the training information set and the verification information set to perform supervised learning and verification on the image processing recognition model until the accuracy of the image processing recognition model meets the model accuracy threshold, thereby obtaining the image processing recognition model.

[0031] Adopting the plurality of sample image information and the plurality of sample equipment status recognition information to construct the image processing recognition model specifically includes: constructing the image processing recognition model based on a convolutional neural network as the basis of the model (the input data at the input end of the image processing recognition model is the real-time video information of the power plant equipment, and the output data at the output end of the image processing recognition model is the equipment status recognition information); dividing the plurality of sample image information and the plurality of sample equipment status recognition information at a ratio of 8:2, decomposing them to obtain a training information set and a verification information set; setting a first model accuracy threshold; inputting the plurality of sample image information and the plurality of sample equipment status recognition information in the training information set as input training data into the convolutional neural network, performing error analysis based on the result of each training and the predicted result (predicted result: the plurality of sample equipment status recognition information), further correcting the weight value and the threshold (by further correcting the weight value and the threshold, the convolutional neural network can be trained until it can be applied to equipment status recognition), and performing supervised learning step by step to obtain a model that can output results consistent with the predicted result. After the output of the model is stabilized (model stability: the output is consistent with the predicted result), input the plurality of sample image information and the plurality of sample equipment status recognition information in the verification information set for verification, calculate the accuracy of the image processing recognition model (image processing recognition model accuracy = number of successful verification times / total number of verification times × 100%), and if the accuracy of the image processing recognition model meets the first model accuracy threshold, it passes the verification, thereby determining the image processing recognition model and providing model support for equipment status recognition.

[0032] Step S231 further includes the following steps.

[0033] S231-1: Install a notation indication recognition algorithm and an indicator light recognition algorithm based on a convolutional neural network as the basis of the model.

[0034] S231-2: Adopt the notation indication recognition algorithm to recognize notation indication information.

[0035] S231-3: Recognize the color of the display light of the enclosure by adopting the display light recognition algorithm.

[0036] S231-4: Recognize and output the equipment status recognition information according to the notation indication information and the color of the display light of the enclosure.

[0037] Specifically, the patrol inspection camera captures various instrument images (such as voltmeters, ammeters, digital display instruments, and knife gate instruments) from within the target power plant. In the process of analyzing the instrument images, instrument digital recognition belongs to scene character recognition. Scene character recognition is carried out in a natural scene and has characteristics such as a relatively complex background, diverse fonts, variable sizes, variable shapes, and irregular directions, and the difficulty of recognition is high.

[0038] When installing the display light recognition algorithm based on the convolutional neural network model, the color of the display light of the enclosure corresponds one-to-one with the display state of the enclosure. According to the operation manual of the main equipment of the display light of the enclosure, query the color of the display light of the enclosure to confirm the display state of the enclosure, modify the double param2 (param2 is a variable in the function, double: double-precision floating-point type) parameter in the HoughCircles() (HoughCircles is a function name) function, reduce the circularity standard of recognition, process the display light area of this part independently, and gradually reduce the magnitude of the value of this parameter according to the idea of the iterative method until a circle is recognized. If the value of param2 is set too low from the beginning, many non-existent circles may be detected, causing a misrecognition phenomenon and resulting in the output of the states of display lights of different colors, so this should not be the case.

[0039] When installing a notation recognition algorithm based on a convolutional neural network as the basis of the model, noise is removed from the input data of the image processing recognition model to provide a clear image for the accurate extraction of the center line of the subsequent pointer. A smoothing window is set (if a smoothing window that is too large or too small is adopted, details of the instrument image, such as boundary contours and lines, will become blurred or the boundary contours will be broken. In the embodiments of the present application, a 5×5 square smoothing window is adopted), and image noise is removed by the median filter of the smoothing window, so that the purpose of noise removal can be achieved, and it is also possible to retain the detailed information in the image, and the pointer is positioned. Generally, the pointer in the instrument image has the characteristics that the upper end is thin, the lower end is thick, and the gray scale is symmetric with respect to the center line, and the center line of the pointer must pass through the rotation axis. The positioning of the pointer can be realized by extracting the center line of the pointer passing through the rotation axis of the dial. To extract the center line of the pointer, a straight line extraction method can be used. The recognition of the indication of the pointer instrument is to recognize the indication of the instrument by using the relationship between the angle of the center line of the pointer with respect to the initial scale of the measurement range of the instrument and the measurement range.

[0040] Taking the pressure gauge dial as an example, its measurement range is (Min, Max). To calculate the indication of the dial, a rectangular coordinate system XY with the rotation axis of the dial as the origin and a coordinate system UV in the image space are constructed. The image pixel coordinate system UV is converted into the dial coordinate system XY. The coordinate origin O is located at the pixel point (m, n) of the rotation axis of the dial. OF is the half-line where the pointer of the dial is located. OI is the half-line from the rotation axis of the dial to the initial scale of the dial. OA is the half-line from the rotation axis of the dial to the maximum scale of the dial. The range of the values of the angles rotated from the negative direction of the X-axis clockwise to the half-lines OA, OI, and OF is (-180°, 180°). The angle rotated clockwise is a positive angle, and the angle rotated counterclockwise is a negative angle.

[0041] When recognizing the notation display information by adopting the notation display recognition algorithm, the notation display information includes a character state, a color state, and a pointer direction. The color of the display lamp of the enclosure is recognized by adopting the display lamp recognition algorithm, and according to the notation display information and the color of the display lamp of the enclosure, the device state is recognized and output according to the usage instructions of the main body equipment of the display lamp of the enclosure, and technical support is provided to ensure the accuracy of the equipment state recognition.

[0042] S30: Connect to the target power plant equipment and obtain the upload data.

[0043] S40: Based on the equipment state recognition information, perform equipment state prediction according to the upload data and obtain the equipment state prediction result.

[0044] Step S40 includes the following steps.

[0045] S41: Construct an equipment state prediction chain.

[0046] S42: Using the equipment state recognition information as the current state, input the upload data into the equipment state prediction chain and perform equipment state prediction.

[0047] S43: Predict and obtain the equipment state prediction result by the equipment state prediction chain.

[0048] Specifically, it is connected to the target power plant equipment by means of communication interconnection to obtain upload data. The upload data includes control commands for which the system has not responded (during the execution of the response to the multiple parallel control commands, the unresponded control commands are stored in the register). Based on the equipment status recognition information, equipment status prediction is performed using the upload data to obtain an equipment status prediction result. Specifically, constructing an equipment status prediction chain, using the Markov chain as the basis of the model, taking the equipment status recognition information as the current state, inputting the upload data into the equipment status prediction chain, predicting the next equipment state (the next equipment state after the current state), obtaining the equipment status prediction result, and providing a reference for subsequent substitution into the equipment status prediction chain.

[0049] Step S41 includes the following steps.

[0050] S411: Construct the equipment status prediction chain using the Markov chain as the basis of the model.

[0051] S412: Set multiple sample equipment status recognition information as state information with multiple sample upload data as reward information, and construct an initialized equipment status prediction chain.

[0052] S413: Randomly extract data from the multiple sample equipment status recognition information and the multiple sample equipment status recognition information to construct a verification information set.

[0053] S414: Set a second model accuracy threshold.

[0054] S415: Verify the initialized equipment status prediction chain using the verification information set until the accuracy of the initialized equipment status prediction chain meets the model accuracy threshold, and obtain the equipment status prediction chain.

[0055] Based on the Markov chain as the basis of the model, constructing an equipment status prediction chain specifically includes using the Markov chain as the basis of the model, setting multiple sample upload data as reward information, setting multiple sample equipment status recognition information as status information, and constructing an initialized equipment status prediction chain; randomly extracting data (random extraction is a prior art) from the multiple sample equipment status recognition information and the multiple sample equipment status recognition information at a ratio of 10% to construct a verification information set; setting a second model accuracy threshold; verifying the initialized equipment status prediction chain by the multiple sample equipment status recognition information and the multiple sample equipment status recognition information in the verification information set until the accuracy of the initialized equipment status prediction chain (prediction chain accuracy = number of successful verifications / total number of verifications × 100%) meets the model accuracy threshold, obtaining the equipment status prediction chain, and providing support for the model for equipment status prediction.

[0056] S50: Set an inspection tour task according to the equipment status prediction result.

[0057] S60: Cause the inspection tour camera to perform an automatic inspection tour on the target power plant equipment according to the inspection tour task.

[0058] As shown in Figure 2, step S50 includes the following steps.

[0059] S51: Obtain an inspection tour route plan according to the equipment status prediction result.

[0060] S52: Obtain an initialized inspection tour route plan during the process of the inspection tour camera's tour.

[0061] S53: Modify the initialized inspection tour route plan according to the inspection tour route plan to obtain a modified inspection tour route plan.

[0062] S54: Add the modified inspection tour route plan to the inspection tour task to update the inspection tour route.

[0063] Setting the patrol inspection task according to the equipment status prediction result specifically includes: obtaining a patrol inspection route plan according to the equipment status prediction result (for the indicator lights and pointer instruments of the same equipment, priority is given to packaging processing, and for those with a large difference between the equipment status prediction result and the equipment status recognition information, priority is given to conducting patrol inspections. A large difference between the equipment status prediction result and the equipment status recognition information means that there may be a large jump in the data and a risk of abnormal operation of the power plant equipment), obtaining an initialized patrol inspection route plan (the initialized patrol inspection route plan is set so that the patrol inspection distance of the initialized patrol inspection route plan is the shortest along the optimization of the patrol inspection distance) during the process of the patrol inspection camera's patrol, modifying the initialized patrol inspection route plan according to the patrol inspection route plan (packaging the operation patrol inspections of the same equipment and giving priority to conducting patrol inspections on those power plant equipment in the target power plant equipment that may have a risk of abnormal operation), and obtaining a modified patrol inspection route plan (simply put, taking the coordinates where there may be a risk of abnormal operation of the power plant equipment as (5, 8), and although the starting point is generally defaulted to (1, 1), the patrol inspection starting point can be modified to (5, 8). The modified patrol inspection route plan may conduct patrol inspections starting from (5, 8) and finally patrol the power plant equipment between (1, 1) and (5, 8)), and in order to ensure the rationality of the patrol route of the patrol inspection camera, adding the modified patrol inspection route plan to the patrol inspection task and performing patrol inspection route update.

[0064] As described above, the method and system for operation patrol inspection and trend analysis of power plant equipment provided by the embodiments of the present application have the following technical effects.

[0065] 1. In the process of the patrol camera's patrol, the video capture equipment is activated to obtain the real-time video information of the power plant equipment, the real-time video information of the power plant equipment is input into the image processing recognition model to obtain the equipment status recognition information, connected to the target power plant equipment, obtain the upload data, based on the equipment status recognition information, perform equipment status prediction by the upload data, obtain the equipment status prediction result, set the patrol inspection task, and cause the patrol camera to perform automatic patrol inspection on the target power plant equipment according to the patrol inspection task. By providing a method and system for the operation patrol inspection and trend analysis of power plant equipment, a recognition model is constructed to improve the accuracy, the patrol inspection route in the patrol inspection task of the patrol camera is optimized using the equipment status prediction, and automatic patrol inspection is performed according to the patrol inspection task, so that the technical effect of improving the operation patrol inspection efficiency and maintaining the reliability of the operation patrol inspection information can be realized.

[0066] 2. According to the equipment status prediction result, obtain the patrol inspection route plan. In the process of the patrol camera's patrol, obtain the initial patrol inspection route plan, modify the initial patrol inspection route plan to obtain the modified patrol inspection route plan and add it to the patrol inspection task, and perform patrol inspection route update. Therefore, the rationality of the patrol route of the patrol camera is guaranteed.

[0067] Example 2

[0068] The embodiment of the present application provides a system for the operation patrol inspection and trend analysis of power plant equipment as shown in FIG. 3 based on the same inventive concept as the method for the operation patrol inspection and trend analysis of power plant equipment according to the foregoing embodiment. The system includes

[0069] A video information acquisition module 100 for activating the video capture equipment to obtain the real-time video information of the power plant equipment in the process of the patrol camera's patrol;

[0070] An equipment status recognition module 200 for inputting the real-time video information of the power plant equipment into the image processing recognition model to obtain the equipment status recognition information;

[0071] An upload data acquisition module 300 that connects to the target power plant facility to acquire upload data,

[0072] A facility state prediction module 400 that performs a facility state prediction based on the facility state recognition information and acquires a facility state prediction result using the upload data,

[0073] A patrol inspection task setting module 500 that sets a patrol inspection task based on the facility state prediction result,

[0074] An automatic patrol inspection module 600 that causes the patrol inspection camera to perform an automatic patrol inspection on the target power plant facility according to the patrol inspection task. The system further includes:

[0075] A sample image information extraction module for synchronously extracting a plurality of sample image information at a previous time when the target power plant facility is in a different facility state,

[0076] A sample facility state recognition information specifying module for specifying corresponding plurality of sample facility state recognition information based on the plurality of sample image information,

[0077] A processing recognition model construction module for constructing the image processing recognition model by adopting the plurality of sample image information and the plurality of sample facility state recognition information,

[0078] A facility state recognition information acquisition module for inputting the real-time video information of the power plant facility into the image processing recognition model and acquiring the facility state recognition information. The system further includes:

[0079] A facility state recognition information acquisition module for inputting the real-time video information of the power plant facility into the image processing recognition model and acquiring the facility state recognition information.

[0080] The system further includes:

[0081] An image processing recognition model construction module for constructing the image processing recognition model based on a convolutional neural network, wherein the input data of the image processing recognition model is real-time video information of power plant facilities, and the output data is equipment status recognition information.

[0082] An information set decomposition module for adopting the plurality of sample image information and the plurality of sample equipment status recognition information, and decomposing them to obtain a training information set and a verification information set.

[0083] A first model accuracy threshold setting module for setting a first model accuracy threshold.

[0084] An image processing recognition model acquisition module for performing supervised learning and verification on the image processing recognition model until the accuracy of the image processing recognition model meets the model accuracy threshold by using the training information set and the verification information set, thereby obtaining the image processing recognition model.

[0085] The system further includes

[0086] A recognition algorithm setting module for installing a notation indication recognition algorithm and an indicator light recognition algorithm based on a convolutional neural network.

[0087] A notation indication information recognition module for recognizing notation indication information by adopting the notation indication recognition algorithm.

[0088] An indicator light color recognition module for the enclosure for recognizing the color of the indicator light of the enclosure by adopting the indicator light recognition algorithm.

[0089] An equipment status recognition information output module for recognizing and outputting the equipment status recognition information based on the notation indication information and the color of the indicator light of the enclosure.

[0090] The system further includes

[0091] a prediction chain construction module for constructing an equipment status prediction chain,

[0092] an equipment status prediction module for inputting the upload data into the equipment status prediction chain with the equipment status recognition information as the current status to perform equipment status prediction,

[0093] and an equipment status prediction result acquisition module for predicting and acquiring an equipment status prediction result according to the equipment status prediction chain.

[0094] The system further includes

[0095] an equipment status prediction chain construction module for constructing the equipment status prediction chain based on a Markov chain as a model basis,

[0096] an initialization equipment status prediction chain construction module for setting multiple sample upload data as reward information, setting multiple sample equipment status recognition information as status information, and constructing an initialization equipment status prediction chain,

[0097] a data random extraction module for randomly extracting data from the multiple sample equipment status recognition information and the multiple sample equipment status recognition information to construct a verification information set,

[0098] a second model accuracy threshold setting module for setting a second model accuracy threshold,

[0099] and an equipment status prediction chain acquisition module for verifying the initialization equipment status prediction chain by the verification information set until the initialization equipment status prediction chain accuracy meets the model accuracy threshold, and acquiring the equipment status prediction chain.

[0100] The system further includes

[0101] An inspection route plan acquisition module for acquiring an inspection route plan based on the equipment status prediction result,

[0102] An initialization inspection route plan acquisition module for acquiring an initialization inspection route plan during the process of the inspection camera's patrol,

[0103] A modified inspection route plan acquisition module for modifying the initialization inspection route plan according to the inspection route plan and acquiring a modified inspection route plan,

[0104] And an inspection route update module for adding the modified inspection route plan to the inspection task to perform inspection route update.

[0105] Any step of the above method may be stored in a non-limiting computer memory as computer instructions or programs, and may be called and recognized by a non-limiting computer processor to implement any method of the embodiments of the present application, and no additional limitation is imposed here.

[0106] Furthermore, the above-mentioned first or second not only represents an order relationship, but may also represent a certain specific concept and / or may represent that individual or all of a plurality of elements may be selected. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Therefore, if these modifications and changes to the present application are within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Description of Reference Signs

[0107] Video information acquisition module 100, equipment status recognition module 200, upload data acquisition module 300, equipment status prediction module 400, inspection task setting module 500, automatic inspection module 600.

Claims

1. A method for operation patrol inspection and trend analysis of power plant equipment, wherein the method is applied to a system for operation patrol inspection and trend analysis, and the system for operation patrol inspection and trend analysis and the patrol inspection camera are communicatively interconnected, and the method includes: During the process of the patrol inspection camera's patrol, starting the video capture equipment to obtain real-time video information of the power plant equipment; Inputting the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment status recognition information; Connecting to the target power plant equipment to obtain upload data; Based on the equipment status recognition information, predicting the equipment status by the upload data to obtain an equipment status prediction result; Setting a patrol inspection task according to the equipment status prediction result; Causing the patrol inspection camera to perform an automatic patrol inspection on the target power plant equipment according to the patrol inspection task; A method for operation patrol inspection and trend analysis of power plant equipment, characterized by including the above.

2. Inputting the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment status recognition information includes: Previously, when the target power plant equipment is in different equipment statuses, synchronously extracting a plurality of sample image information; Identifying corresponding a plurality of sample equipment status recognition information based on the plurality of sample image information; Constructing the image processing recognition model by adopting the plurality of sample image information and the plurality of sample equipment status recognition information; Inputting the real-time video information of the power plant equipment into the image processing recognition model to obtain the equipment status recognition information; The method according to claim 1, characterized by including the above.

3. Constructing the image processing recognition model by adopting the plurality of sample image information and the plurality of sample equipment status recognition information includes: Constructing the image processing recognition model based on a convolutional neural network as the basis of the model, wherein the input data of the image processing recognition model is the real-time video information of the power plant equipment, and the output data is the equipment status recognition information; Adopting the plurality of sample image information and the plurality of sample equipment status recognition information, decomposing them to obtain a training information set and a verification information set; Setting a first model accuracy threshold; Using the training information set and the verification information set, perform supervised learning and verification on the image processing recognition model until the accuracy of the image processing recognition model meets the model accuracy threshold, thereby obtaining the image processing recognition model. The method according to claim 2, characterized by including the above.

4. Based on a convolutional neural network as the basis of the model, constructing the image processing recognition model further includes: Based on a convolutional neural network as the basis of the model, installing a notation recognition algorithm and an indicator light recognition algorithm. Adopting the notation recognition algorithm to recognize notation information. Adopting the indicator light recognition algorithm to recognize the color of the indicator light of the enclosure. Based on the notation information and the color of the indicator light of the enclosure, recognizing and outputting the equipment status recognition information. The method according to claim 3, characterized by including the above.

5. Performing equipment status prediction based on the uploaded data according to the equipment status recognition information and obtaining an equipment status prediction result includes: Constructing an equipment status prediction chain. Using the equipment status recognition information as the current state, inputting the uploaded data into the equipment status prediction chain to perform equipment status prediction. Predicting and obtaining an equipment status prediction result by the equipment status prediction chain. The method according to claim 1, characterized by including the above.

6. Constructing the equipment status prediction chain includes: Based on a Markov chain as the basis of the model, constructing the equipment status prediction chain. Using a plurality of sample uploaded data as reward information, setting a plurality of sample equipment status recognition information as state information, and constructing an initialized equipment status prediction chain. Randomly extracting data from the plurality of sample equipment status recognition information and the plurality of sample equipment status recognition information to construct a verification information set. Setting a second model accuracy threshold. Using the verification information set to verify the initialized equipment status prediction chain until the accuracy of the initialized equipment status prediction chain meets the model accuracy threshold, and obtaining the equipment status prediction chain. The method according to claim 5, characterized by including the above.

7. Setting an inspection task according to the equipment status prediction result includes: Obtaining an inspection route plan according to the equipment status prediction result. In the process of the patrol inspection camera's patrol, obtaining an initial patrol inspection route plan; According to the patrol inspection route plan, modifying the initial patrol inspection route plan to obtain a modified patrol inspection route plan; Adding the modified patrol inspection route plan to the patrol inspection task to perform patrol inspection route update; The method according to claim 1, characterized by including the above.

8. A system for the operation patrol inspection and trend analysis of power plant equipment, used to implement the method for the operation patrol inspection and trend analysis of power plant equipment according to any one of claims 1 to 7. In the process of the patrol inspection camera's patrol, a video information acquisition module for starting a video capture device to obtain real-time video information of power plant equipment; An equipment status recognition module for inputting the real-time video information of the power plant equipment into an image processing recognition model to obtain equipment status recognition information; An upload data acquisition module for connecting to the target power plant equipment to obtain upload data; An equipment status prediction module for performing equipment status prediction based on the upload data according to the equipment status recognition information to obtain an equipment status prediction result; A patrol inspection task setting module for setting a patrol inspection task according to the equipment status prediction result; An automatic patrol inspection module for causing the patrol inspection camera to perform an automatic patrol inspection on the target power plant equipment according to the patrol inspection task; A system for the operation patrol inspection and trend analysis of power plant equipment, characterized by comprising the above.

Images