Power transmission and transformation project progress identification and analysis method and system based on deep learning

Abstract

The invention discloses a deep learning-based transmission and transformation project progress identification and analysis method and system, and the method comprises the steps: firstly collecting and preprocessing the multimedia data of a transmission and transformation site, constructing a project progress data set, training a deep learning model through a plane graph and the progress data set, inputting new progress data into the model, and identifying a current project state; comparing the plan progress, analyzing the deviation, and finally performing comprehensive trend prediction based on the model and the key influence factors. In the application of the design, key factors are obtained through a recursive transformer fusion network model fusing various modules and various differential data, and comprehensive trend analysis is realized in combination with differences of different voltages, so that deep-level features can be effectively extracted from various data sources, high-precision identification of the project progress is realized, and the accuracy of the project progress is improved. Compared with a traditional manual checking and recording mode, the method has the advantages that the recognition efficiency and accuracy are greatly improved, the interference of human factors is reduced, and the real-time performance and reliability of progress recognition are ensured.

Description

Technical Field

[0001] The present invention relates to a progress analysis method, and belongs to the field of power transmission and transformation project progress, and in particular to a power transmission and transformation project progress identification and analysis method and system based on deep learning. Background Art

[0002] In the power system, power transmission and transformation projects are the key link to ensure the safe and stable operation of power. With the continuous development of social economy, the demand for electricity has increased significantly, and large-scale power transmission and transformation projects have also increased. However, the traditional progress monitoring and management methods of power transmission and transformation projects have many shortcomings and cannot meet the requirements of modern project management.

[0003] Traditional power transmission and transformation project progress monitoring mainly relies on manual inspection and recording. Manual methods are difficult to process and analyze large amounts of data in real time, and cannot reflect the actual progress of the project in a timely manner, increasing the difficulty and risk of project management. Not only is it inefficient, but it is also easily affected by human factors, and the accuracy of progress identification is difficult to guarantee. Although some existing data analysis methods, such as time series analysis and regression analysis, can play a role in certain specific scenarios, they often show problems such as low prediction accuracy and poor adaptability when dealing with complex and dynamic engineering data. It is difficult to effectively deal with various complex situations in large-scale power transmission and transformation projects, and it is impossible to provide reliable decision support for project management. Therefore, there is an urgent need for a means to improve the accuracy and efficiency of project progress monitoring, achieve scientific prediction and trend analysis of project progress, and provide strong support for project management. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned defects and problems existing in the prior art and to provide an accurate and efficient method and system for identifying and analyzing the progress of power transmission and transformation projects based on deep learning.

[0005] To achieve the above objectives, the technical solution of the present invention is: a method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning, comprising:

[0006] S1. Acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set;

[0007] S2. Based on the plan data and the power transmission and transformation project progress data set, a deep learning model is trained to obtain a deep learning model of the power transmission and transformation project;

[0008] S3, input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status;

[0009] S4. Compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project;

[0010] S5. Conduct trend analysis on the progress of future power transmission and transformation projects based on the deep learning model and key factors of power transmission and transformation projects, and conduct comprehensive trend analysis based on the differences in construction sites of different voltage levels and trend analysis results to obtain comprehensive trend analysis results.

[0011] The step S1 specifically includes:

[0012] S11, obtaining CAD plan data of the power transmission and transformation project site, and photographing the power transmission and transformation project site to obtain image data and video data of the power transmission and transformation project site;

[0013] The image data includes a static image, and the video data includes a continuous video frame;

[0014] S12, performing denoising and image enhancement processing on the image data, and annotating the image data to form an image data set; the expression of the image data set is as follows:

[0015]

[0016] Where: D img (i) is the image dataset, I u (x1, y1, h, θ, t1), are image data collected by different camera devices. x1, y1, x2, y2 are the spatial coordinates of the image, h and z are the shooting heights, θ, ψ are all shooting angles, t1 and t2 are all time stamps;

[0017] S13, extracting frames from the video data and marking them to form a video data set; the expression of the video data set is as follows:

[0018] D vid (i) = {V C (x3, y3, α, β, t3)};

[0019] Where: D vid (i) is the video dataset, V C (x3, y3, α, β, t3) is the video data collected by the camera device, x3, y3 are the spatial coordinates of the video frame, α, β are the shooting angles, and t3 is the timestamp of the video frame;

[0020] S14. Classify and store the image data set and the video data set to form a power transmission and transformation project progress data set.

[0021] The step S2 specifically includes:

[0022] S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module;

[0023] S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows:

[0024]

[0025] Where: S out For spatial feature output, Transformer spatial is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s,k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s 、b t are all offset items, DCAD(i, k) is the CAD plan data, D final (i, l) is the data set of power transmission and transformation project progress;

[0026] S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows:

[0027] M fused =λ1·S out +λ2·T out +λ3·Attn(W f ·[S out ; T out ]+b f );

[0028] Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term;

[0029] S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows:

[0030]

[0031] in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module;

[0032] S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows:

[0033]

[0034] Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term;

[0035] S26. Input the adjusted weight features into the progress feature classification module to generate progress features of the power transmission and transformation project at different stages, complete the training of the deep learning model, and obtain the deep learning model of the power transmission and transformation project;

[0036] The expressions of the progress characteristics of the substation engineering at different stages are as follows:

[0037] P stage (i) = Softmax(W c ·W adjusted +b c );

[0038] Where: P stage (i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b C is the classifier bias.

[0039] The step S4 specifically includes:

[0040] S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows:

[0041]

[0042] Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan(i, j) is the jth item of the actual progress characteristic of the power transmission and transformation project, w j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function;

[0043] S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows:

[0044]

[0045] Where: D lag (i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and β j is the weight of the corresponding item, Sigmoid is the S-type activation function;

[0046] S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows:

[0047]

[0048] Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, Yj is the weight of the corresponding item, and Tanh is the hyperbolic tangent function;

[0049] S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows:

[0050]

[0051] Among them: A finance (i) is the progress analysis result combined with financial data, F budget (i, j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the rectified linear unit function with leakage;

[0052] S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows:

[0053]

[0054] Where: K factor(i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Softmax is the normalized exponential function.

[0055] The step S5 specifically includes:

[0056] S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows:

[0057] P future (i, t) = M trained (D final (i), t);

[0058] P weighted (i, t) = P future (i, t)×K factor (i);

[0059] Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t;

[0060] S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows:

[0061]

[0062] Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period;

[0063] S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows:

[0064]

[0065] Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, V) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items;

[0066] S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed; the expression of the comprehensive trend analysis is as follows:

[0067] T comprehensive (i) = α·T trend (i)+β·D compare (i, v);

[0068] Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

[0069] A power transmission and transformation project progress identification and analysis system based on deep learning, the system comprising:

[0070] The data set acquisition module is used to acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set;

[0071] A deep learning model training module is used to train a deep learning model based on the floor plan data and the power transmission and transformation project progress data set to obtain a deep learning model for the power transmission and transformation project;

[0072] The progress status acquisition module is used to input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status;

[0073] The key factor identification module is used to compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project;

[0074] The comprehensive trend analysis module is used to conduct trend analysis on the progress of future power transmission and transformation projects based on the deep learning model and key factors of power transmission and transformation projects, and to conduct comprehensive trend analysis based on the differences in construction sites of different voltage levels and the trend analysis results to obtain comprehensive trend analysis results.

[0075] The deep learning model training module is trained according to the following steps:

[0076] S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module;

[0077] S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows:

[0078]

[0079] Where: S out For spatial feature output, Transformer spatial is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s,k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s , b t are all bias terms, D CAD (i, k) is the CAD plan data, D final (i, l) is the data set of power transmission and transformation project progress;

[0080] S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows:

[0081] M fused =λ1·S out +λ2·T out +λ3·Attn(W f ·[S out ; T out ]+b f );

[0082] Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term;

[0083] S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows:

[0084]

[0085] in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module;

[0086] S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows:

[0087]

[0088] Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term;

[0089] S26. Input the adjusted weight features into the progress feature classification module to generate progress features of the power transmission and transformation project at different stages, complete the training of the deep learning model, and obtain the deep learning model of the power transmission and transformation project;

[0090] The expressions of the progress characteristics of the substation engineering at different stages are as follows:

[0091] P stage (i) = Softmax(W c ·W adjusted +b c );

[0092] Where: P stage (i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b c is the classifier bias.

[0093] The key factor identification module identifies key factors according to the following steps:

[0094] S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows:

[0095]

[0096] Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan (i, j) is the jth item of the actual progress characteristic of the power transmission and transformation project, w j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function;

[0097] S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows:

[0098]

[0099] Where: D lag(i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and β j is the weight of the corresponding item, Sigmoid is the S-type activation function;

[0100] S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows:

[0101]

[0102] Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, Yj is the weight of the corresponding item, and Tanh is the hyperbolic tangent function;

[0103] S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows:

[0104]

[0105] Among them: A finance (i) is the progress analysis result combined with financial data, F budget (i, j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the rectified linear unit function with leakage;

[0106] S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows:

[0107]

[0108] Where: K factor (i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Softmax is the normalized exponential function.

[0109] The comprehensive trend analysis module analyzes the comprehensive trend according to the following steps:

[0110] S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows:

[0111] P future(i, t) = M trained (D final (i), t);

[0112] P weighted (i, t) = P future (i, t)×K factor (i);

[0113] Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t;

[0114] S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows:

[0115]

[0116] Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period;

[0117] S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows:

[0118]

[0119] Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, v) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items;

[0120] S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed; the expression of the comprehensive trend analysis is as follows:

[0121] T comprehensive (i) = α·T trend (i)+β·D compare (i, v);

[0122] Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

[0123] A power transmission and transformation project progress identification and analysis device based on deep learning, the device comprising a processor and a memory;

[0124] The memory is used to store computer program code and transmit the computer program code to the processor;

[0125] The processor is used to execute the above-mentioned power transmission and transformation project progress identification and analysis method based on deep learning according to the instructions in the computer program code.

[0126] Compared with the prior art, the present invention has the following beneficial effects:

[0127] The present invention discloses a method and system for identifying and analyzing the progress of a power transmission and transformation project based on deep learning. The method first collects and preprocesses multimedia data from a power transmission and transformation site to construct a project progress data set, then uses a floor plan and a progress data set to train a deep learning model, then inputs the new progress data into the model to identify the current project status, compares the planned progress, analyzes the deviation, and finally performs a comprehensive trend prediction based on the model and key influencing factors. In the application of this design, a recursive transformer fusion network model that integrates multiple modules and multiple difference data are used to obtain key factors, and a comprehensive trend analysis is achieved in combination with the differences in different voltages, so that deep-level features can be effectively extracted from multiple data sources, and high-precision identification of the project progress is achieved. Compared with traditional manual inspection and recording methods, the recognition efficiency and accuracy are greatly improved, the interference of human factors is reduced, and the real-time and reliability of progress identification are ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0128] Figure 1 It is a flow chart of the steps of the method of the present invention.

[0129] Figure 2 It is a structural diagram of the recursive transformer fusion network model in the present invention.

[0130] Figure 3 It is a schematic diagram of the system structure of the present invention.

[0131] Figure 4 It is a schematic diagram of the device structure of the present invention.

[0132] In the figure: data set acquisition module 1, deep learning model training module 2, progress status acquisition module 3, key factor identification module 4, comprehensive trend analysis module 5, processor 6, memory 7, computer program code 71. DETAILED DESCRIPTION

[0133] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0134] Embodiment 1:

[0135] See also Figure 1 , a method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning, comprising:

[0136] S1. Acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set;

[0137] Furthermore, the step S1 specifically includes:

[0138] S11, obtaining CAD plan data of the power transmission and transformation project site, and photographing the power transmission and transformation project site to obtain image data and video data of the power transmission and transformation project site;

[0139] In this embodiment, a camera or video equipment can be used to shoot the power transmission and transformation project site from multiple angles and multiple heights to obtain image data and video data of the power transmission and transformation project site;

[0140] In this embodiment, the camera device or video device includes a drone, a surveillance camera, and a robot dog device. When used, the drone device is used to shoot the power transmission and transformation project site from different heights and angles in the air, the surveillance camera device at a fixed position is used to continuously shoot the power transmission and transformation project site from different angles and positions on the ground, and the mobile device robot dog is used to shoot the power transmission and transformation project site from different positions on the ground.

[0141] The image data includes a static image, and the video data includes a continuous video frame;

[0142] The objects photographed include roads, squares, walls, gates, drainage wells, main transformers, voltage transformers, circuit breakers, reactors, low-voltage capacitors, secondary cabinets, and cables;

[0143] S12, performing denoising and image enhancement processing on the image data, and annotating the image data to form an image data set; the expression of the image data set is as follows:

[0144]

[0145] Where: D img (i) is the image dataset, I u (x1, y1, h, θ, t1), is the image data collected by different photographic equipment (different photographic equipment such as drones and robot dogs), x1, y1, x2, y2 are the spatial coordinates of the image, h, z are the shooting heights, θ, ψ are all shooting angles, t1 and t2 are all time stamps;

[0146] S13, extracting frames from the video data and marking them to form a video data set; the expression of the video data set is as follows:

[0147] D vid (i) = {V C (x3, y3, α, β, t3)};

[0148] Where: D vid (i) is the video dataset, V C (x3, y3, α, β, t3) is the video data collected by the camera (such as a surveillance camera), x3, y3 are the spatial coordinates of the video frame, α, β are the shooting angles, and t3 is the timestamp of the video frame;

[0149] S14. Classify and store the image data set and the video data set to form a power transmission and transformation project progress data set.

[0150] S2. Based on the plan data and the power transmission and transformation project progress data set, a deep learning model is trained to obtain a deep learning model of the power transmission and transformation project;

[0151] Furthermore, the step S2 specifically includes:

[0152] S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module;

[0153] S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows:

[0154]

[0155] Where: S out For spatial feature output, Transformer spatidl is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s,k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s , b t are all bias terms, D CAD (i, k) is the CAD plan data, D final (i, l) is the data set of power transmission and transformation project progress;

[0156] S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows:

[0157] M fused =λ1·S out +λ2·T out +λ3·Attn(W f ·[S out ; T out ]+b f );

[0158] Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term;

[0159] S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows:

[0160]

[0161] in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module;

[0162] S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows:

[0163]

[0164] Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term;

[0165] S26. Input the adjusted weight features into the progress feature classification module to generate progress features of the power transmission and transformation project at different stages, complete the training of the deep learning model, and obtain the deep learning model of the power transmission and transformation project;

[0166] The expressions of the progress characteristics of the substation engineering at different stages are as follows:

[0167] Pstage (i) = Softmax(W c ·W adjusted +b c );

[0168] Where: P stage (i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b c is the classifier bias.

[0169] S3, input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status;

[0170] Furthermore, the step S3 specifically refers to:

[0171] Based on the deep learning model of power transmission and transformation engineering, the newly collected image data and video data are formed into a new power transmission and transformation engineering progress data set and input into the model for progress identification. By identifying the status of station area buildings, main production buildings, distribution device buildings, auxiliary production buildings, fire protection systems, and electrical equipment, the current progress status of the power transmission and transformation project is obtained.

[0172] S4. Compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project;

[0173] Furthermore, the step S4 specifically includes:

[0174] S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows:

[0175]

[0176] Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan (i, j) is the jth item of the actual progress characteristic of the power transmission and transformation project, w j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function, which is used to filter positive differences;

[0177] S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows:

[0178]

[0179] Where: D lag (i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and β j is the weight of the corresponding item, Sigmoid is the S-type activation function used to filter the lag and lead differences;

[0180] S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows:

[0181]

[0182] Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, γ j is the weight of the corresponding item, Tanh is the hyperbolic tangent function, which is used for smooth difference analysis;

[0183] S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows:

[0184]

[0185] Among them: A finance (i) is the progress analysis result combined with financial data, F budget ( i , j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the leaky rectified linear unit function used to screen financial impact;

[0186] S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows:

[0187]

[0188] Where: K factor (i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Softmax is a normalized exponential function used to determine the influence of key factors.

[0189] S5. Conduct trend analysis on the progress of future power transmission and transformation projects based on the deep learning model and key factors of power transmission and transformation projects, and conduct comprehensive trend analysis based on the differences in construction sites of different voltage levels and trend analysis results to obtain comprehensive trend analysis results.

[0190] Furthermore, the step S5 specifically includes:

[0191] S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows:

[0192] P future (i, t) = M trained (D final (i), t);

[0193] P weighted (i, t) = P future (i, t)×K factor (i);

[0194] Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t;

[0195] S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows:

[0196]

[0197] Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period;

[0198] S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows:

[0199]

[0200] Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, v) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items;

[0201] S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed; the expression of the comprehensive trend analysis is as follows:

[0202] T comprehensive (i) = α·T trend (i)+β·D compare (i, v);

[0203] Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

[0204] In the application, the comprehensive trend analysis results reflect the trend of future project progress and can help the project management team predict the project's funding needs and investment returns. For example, if the comprehensive trend analysis results show that the project progress at a certain stage may lag behind, it can be foreseen that additional funds or resources are needed to catch up with the planned progress; if it is shown to be ahead of schedule, it can be known that the current funds or resources can meet the current progress requirements. Therefore, comprehensive trend analysis can directly affect investment decisions and provide reference for fund allocation, risk control and resource scheduling by generating investment statistical reports.

[0205] The relationship between progress analysis and investment statistics is that the progress of the project directly affects the efficiency and effect of the investment. For example, at different stages of the project, the project management team can adjust the fund allocation strategy according to the actual progress of the project; delayed or advanced progress may lead to the reallocation of funds, affecting the budget and fund use of the entire project; therefore, progress analysis provides basic data for investment statistics, helping to ensure that the capital investment matches the progress of the project, thereby optimizing the return on investment and reducing financial risks.

[0206] In this embodiment, the scenario is preset as a 110kV substation construction project undertaken by a provincial power company. The project involves multiple engineering progress nodes such as station area buildings, main production buildings, distribution device buildings, auxiliary production buildings, fire protection systems and electrical equipment. The project is located in complex terrain, with changeable climate and huge engineering volume. Traditional progress monitoring methods are difficult to meet the needs of project management.

[0207] In this scenario, drones, surveillance cameras and robot dogs were first used to photograph the substation construction site from multiple angles and positions, collecting a large amount of high-quality image and video data. The collected data included engineering elements such as roads and squares inside and outside the station, walls, gates, drainage wells, main transformers, voltage transformers, circuit breakers, reactors, low-voltage capacitors, secondary panels and cables. The precise selection and collection of data objects ensured the comprehensiveness and real-time nature of the data, providing a solid data foundation for subsequent analysis.

[0208] In the data preprocessing stage, we performed denoising, image enhancement and frame extraction on the collected image and video data to form high-quality image and video datasets; the specific steps include using image denoising algorithms to remove noise in images and using image enhancement algorithms to improve image clarity and ensure data reliability.

[0209] Then, see Figure 2 , the denoised and enhanced image data and the extracted video frame data were annotated to form a data set of power transmission and transformation project progress. Then, we used the CAD floor plan data and the data set of power transmission and transformation project progress to train the deep learning model; the recursive transformer fusion network model used combines the multimodal fusion module, the dynamic self-attention adjustment module, the recursive transformer module and the progress feature classification module. By extracting and fusing spatial features and temporal features, high-precision recognition of project progress is achieved. During the model training process, the annotated progress data set is used for multi-scale feature fusion and analysis to ensure that the model can capture subtle changes in project progress and improve the comprehensiveness and accuracy of recognition.

[0210] Through the trained deep learning model, the progress of the new data actually collected was identified, and the key nodes and current progress status of the project were automatically detected and identified. Subsequently, the actual progress characteristics identified were compared with the planned progress characteristics, and the lagging or advanced parts of the project progress were analyzed. Combined with the work tickets, financial budget and expenditure data, a comprehensive analysis was conducted to find out the key factors affecting the project progress.

[0211] In order to verify the effectiveness of the method of the present invention, we selected 10 main progress nodes in the project for comparative analysis; the following Table 1 is the progress identification and analysis data of the method of the present invention and the traditional method at these nodes:

[0212]

[0213] From the data in Table 1, it can be seen that the method of the present invention is significantly superior to the traditional method in terms of recognition accuracy and processing time; in terms of recognition accuracy, the method of the present invention has increased by about 25% on average, and the processing time has been reduced by about 60%. These data prove the efficiency and accuracy of the present invention in the identification and analysis of the progress of large-scale power transmission and transformation projects.

[0214] In addition, through the trend analysis of the deep learning model, we predicted the future project progress, and combined the video image data of the construction sites with different voltage levels to conduct difference analysis and comparison, providing a scientific reference basis for the project's investment statistics; during the implementation of the entire project, the method of the present invention effectively improved the intelligence and automation level of project progress monitoring, reduced manual intervention, reduced the difficulty and risk of project management, and ensured the smooth implementation of the project and its completion on time.

[0215] Embodiment 2:

[0216] See also Figure 3 , a power transmission and transformation project progress identification and analysis system based on deep learning, the system comprising:

[0217] The data set acquisition module 1 is used to acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set;

[0218] Furthermore, the data set acquisition module 1 acquires the data set according to the following steps:

[0219] S11, obtaining CAD plan data of the power transmission and transformation project site, and photographing the power transmission and transformation project site to obtain image data and video data of the power transmission and transformation project site;

[0220] The image data includes a static image, and the video data includes a continuous video frame;

[0221] S12, performing denoising and image enhancement processing on the image data, and annotating the image data to form an image data set; the expression of the image data set is as follows:

[0222]

[0223] Where: D img (i) is the image dataset, I u (x1, y1, h, θ, t1), are image data collected by different camera devices. x1, y1, x2, y2 are the spatial coordinates of the image, h and z are the shooting heights, θ, ψ are all shooting angles, t1 and t2 are all time stamps;

[0224] S13, extracting frames from the video data and marking them to form a video data set; the expression of the video data set is as follows:

[0225] D vid (i) = {V C (x3, y3, α, β, t3)};

[0226] Where: D vid (i) is the video dataset, V C (x3, y3, α, β, t3) is the video data collected by the camera device, x3, y3 are the spatial coordinates of the video frame, α, β are the shooting angles, and t3 is the timestamp of the video frame;

[0227] S14. Classify and store the image data set and the video data set to form a power transmission and transformation project progress data set.

[0228] The deep learning model training module 2 is used to train the deep learning model based on the plan data and the power transmission and transformation project progress data set to obtain the power transmission and transformation project deep learning model;

[0229] Furthermore, the deep learning model training module 2 is trained according to the following steps:

[0230] S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module;

[0231] S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows:

[0232]

[0233] Where: S out For spatial feature output, Transformer spatial is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s,k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s 、b t are all bias terms, D CAD (i, k) is the CAD plan data, D final (i, l) is the data set of power transmission and transformation project progress;

[0234] S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows:

[0235] M fused =λ1·S out +λ2·T out +λ3·Attn(Wf ·[S out ; T ou t]+b f );

[0236] Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term;

[0237] S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows:

[0238]

[0239] in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module;

[0240] S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows:

[0241]

[0242] Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term;

[0243] S26. Input the adjusted weight features into the progress feature classification module to generate progress features of the power transmission and transformation project at different stages, complete the training of the deep learning model, and obtain the deep learning model of the power transmission and transformation project;

[0244] The expressions of the progress characteristics of the substation engineering at different stages are as follows:

[0245] P stage (i) = Softmax(W c ·W adjusted +b c );

[0246] Where: P stage(i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b c is the classifier bias.

[0247] The progress status acquisition module 3 is used to input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status;

[0248] Furthermore, the progress status acquisition module 3 acquires the current progress status in the following manner:

[0249] Based on the deep learning model of power transmission and transformation engineering, the newly collected image data and video data are formed into a new power transmission and transformation engineering progress data set and input into the model for progress identification. By identifying the status of station area buildings, main production buildings, distribution device buildings, auxiliary production buildings, fire protection systems, and electrical equipment, the current progress status of the power transmission and transformation project is obtained.

[0250] The key factor identification module 4 is used to compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project;

[0251] Furthermore, the key factor identification module 4 identifies the key factors according to the following steps:

[0252] S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows:

[0253]

[0254] Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan (i, j) is the jth item of the actual progress characteristic of the power transmission and transformation project, w j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function;

[0255] S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows:

[0256]

[0257] Where: D lag (i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and βj is the weight of the corresponding item, Sigmoid is the S-type activation function;

[0258] S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows:

[0259]

[0260] Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, Yj is the weight of the corresponding item, and Tanh is the hyperbolic tangent function;

[0261] S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows:

[0262]

[0263] Among them: A finance (i) is the progress analysis result combined with financial data, F budget (i, j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the rectified linear unit function with leakage;

[0264] S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows:

[0265]

[0266] Where: K factor (i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Soffmax is the normalized exponential function.

[0267] The comprehensive trend analysis module 5 is used to perform trend analysis on the progress of future power transmission and transformation projects based on the deep learning model and key factors of power transmission and transformation projects, and to perform comprehensive trend analysis based on the differences between construction sites of different voltage levels and the trend analysis results to obtain comprehensive trend analysis results.

[0268] Further, the comprehensive trend analysis module 5 analyzes the comprehensive trend according to the following steps:

[0269] S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows:

[0270] P future (i, t) = M trained (D final (i), t);

[0271] P weighted (i, t) = P future (i, t)×K factor (i);

[0272] Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t;

[0273] S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows:

[0274]

[0275] Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period;

[0276] S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows:

[0277]

[0278] Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, v) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items;

[0279] S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed; the expression of the comprehensive trend analysis is as follows:

[0280] T comprehensive (i) = α·T trend (i)+β·D compare (i, v);

[0281] Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

[0282] Embodiment 3:

[0283] See also Figure 4 , in this embodiment, it also includes a power transmission and transformation project progress identification and analysis device based on deep learning, the device includes a processor 6 and a memory 7;

[0284] The memory 7 is used to store computer program code 71 and transmit the computer program code 71 to the processor 6;

[0285] The processor 6 is used to execute the power transmission and transformation project progress identification and analysis method based on deep learning described in Example 1 according to the instructions in the computer program code 71.

[0286] This embodiment also includes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed on a computer, the method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning described in Example 1 is implemented.

[0287] Generally speaking, the computer instructions for implementing the method of the present invention may be carried in any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media may include any computer-readable media, except for the signal itself that is temporarily propagating.

[0288] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EKROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0289] Computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, SMalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages, in particular, Python suitable for neural network computing and platform frameworks based on TensorFlow, PyTorch, etc. can be used. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer or to an external computer (for example, using an Internet service provider to connect via the Internet) through any type of network, including a local area network (LAN) or a wide area network (WAN).

[0290] The above-mentioned device and non-temporary computer-readable storage medium can be found in the specific description of a method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning and its beneficial effects, which will not be repeated here.

[0291] Although the embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning, characterized in that: include: S1. Acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set; S2. Based on the plan data and the power transmission and transformation project progress data set, a deep learning model is trained to obtain a deep learning model of the power transmission and transformation project; S3, input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status; S4. Compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project; S5. Conduct trend analysis on the progress of future power transmission and transformation projects based on the deep learning model and key factors of power transmission and transformation projects, and conduct comprehensive trend analysis based on the differences in construction sites of different voltage levels and trend analysis results to obtain comprehensive trend analysis results.

2. According to the method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning in claim 1, it is characterized by: The step S1 specifically includes: S11, obtaining CAD plan data of the power transmission and transformation project site, and photographing the power transmission and transformation project site to obtain image data and video data of the power transmission and transformation project site; The image data includes a static image, and the video data includes a continuous video frame; S12, performing denoising and image enhancement processing on the image data, and annotating the image data to form an image data set; the expression of the image data set is as follows: Where: D img (i) is the image dataset, I u (x1, y1, h, θ, t1), are image data collected by different camera equipment. x1, y1, x2, y2 are the spatial coordinates of the image, h and z are the shooting heights, θ, ψ are all shooting angles, t1 and t2 are all time stamps; S13, extracting frames from the video data and marking them to form a video data set; the expression of the video data set is as follows: D vid (i)={V C (x3,y3,α,β,t3)}; Where: D vid (i) is the video dataset, V C (x3, y3, α, β, t3) is the video data collected by the camera device, x3, y3 are the spatial coordinates of the video frame, α, β are the shooting angles, and t3 is the timestamp of the video frame; S14. Classify and store the image data set and the video data set to form a power transmission and transformation project progress data set.

3. The method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning according to claim 1, characterized in that: The step S2 specifically includes: S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module; S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows: Where: S out For spatial feature output, Transformer spatial is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s，k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s 、b t Both are bias terms, D CAD (i, k) is the CAD plan data, D final (i, 1) is the data set of power transmission and transformation project progress; S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows: M fused =λ1·S out +λ2·T out +λ3·Attn(W f ·[S out ;T out ]+b f ); Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term; S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows: in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module; S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows: Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term; S26, inputting the adjusted weight features into the progress feature classification module, generating progress features of the power transmission and transformation project at different stages, completing the training, and obtaining a deep learning model of the power transmission and transformation project; The expressions of the progress characteristics of the substation engineering at different stages are as follows: P stage (i)=Softmax(W c ·W adjusted +b c ); Where: P stage (i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b c is the classifier bias.

4. The method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning according to claim 1, characterized in that: The step S4 specifically includes: S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows: ReLU(P stage (i,j)-P plan (i,j))) Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan (i, j) is the jth item of the actual progress characteristic of the power transmission and transformation project, W j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function; S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows: Where: D lag (i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and β j is the weight of the corresponding item, Sigmoid is the S-type activation function; S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows: Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, Yj is the weight of the corresponding item, and Tanh is the hyperbolic tangent function; S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows: LeakyReLU(D diff (i,j))) Among them: A finance (i) is the progress analysis result combined with financial data, F budget (i, j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the rectified linear unit function with leakage; S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows: Softmax(A task (i,j)+A finance (i,j))) Where: K factor (i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Softmax is the normalized exponential function.

5. The method for identifying and analyzing the progress of power transmission and transformation projects based on deep learning according to claim 1, characterized in that: The step S5 specifically includes: S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows: P future (i,t)=M trained (D final (i),t); P weighted (i,t)=P future (i,t)×K factor (i); Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t; S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows: Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period; S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows: Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, v) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items; S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed to obtain a comprehensive trend analysis result. The expression of the comprehensive trend analysis is as follows: T comprehensive (i)=α·T trend (i)+β·D compare (i,v); Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

6. A power transmission and transformation project progress identification and analysis system based on deep learning, characterized in that: The system comprises: The data set acquisition module (1) is used to acquire the plan data, image data and video data of the power transmission and transformation project site, and pre-process the image data and video data to obtain the power transmission and transformation project progress data set; A deep learning model training module (2) is used to train a deep learning model based on the plan data and the power transmission and transformation project progress data set to obtain a power transmission and transformation project deep learning model; A progress status acquisition module (3) is used to input the newly acquired power transmission and transformation project progress data set into the power transmission and transformation project deep learning model to obtain the current progress status; The key factor identification module (4) is used to compare the current progress status with the actual progress of the power transmission and transformation project, analyze the progress difference of the lagging or leading parts, and obtain the key factors affecting the progress of the power transmission and transformation project; The comprehensive trend analysis module (5) is used to perform trend analysis on the progress of future power transmission and transformation projects based on the deep learning model of power transmission and transformation projects and key factors, and to perform comprehensive trend analysis based on the differences between construction sites of different voltage levels and the trend analysis results to obtain comprehensive trend analysis results.

7. The power transmission and transformation project progress identification and analysis system based on deep learning according to claim 6 is characterized by: The deep learning model training module (2) is trained according to the following steps: S21, using the plan data and the power transmission and transformation project progress data set as input data for the deep learning model; the deep learning model is a recursive transformer fusion network model, which includes a multimodal fusion module, a dynamic self-attention adjustment module, a recursive transformer module, and a progress feature classification module; S22. In the recursive transformer fusion network model, a dual-path network architecture including a spatial feature extraction path and a temporal feature extraction path is constructed; the expression of the dual-path network architecture is as follows: Where: S out For spatial feature output, Transformer spatial is the transformer module of spatial features, T out For temporal feature output, Transformer temporal is the transformer module of time characteristics, W s,k , W t，l are the weights of the spatial path and the temporal path, respectively, and b s 、b t Both are bias terms, D CAD (i, k) is the CAD plan data, D final (i, l) is the data set of power transmission and transformation project progress; S23, based on the multimodal fusion module, the spatial feature and the temporal feature are fused to generate a multimodal feature; the expression of the multimodal feature is as follows: M fused =λ1·S out +λ2·T out +λ3·Attn(W f ·[S out ;T out ]+b f ); Where: M fused is the multimodal feature output, λ1, λ2, λ3 are dynamic weight coefficients, Attn is the self-attention mechanism, W f is the fusion weight matrix, [S out ; T out ] is feature splicing, b f is the fusion bias term; S24. Based on the recursive transformer module, a recursive transformer operation is used to perform deep recursive feature extraction on the multimodal features; the expression of the recursive feature is as follows: in: is the recursive feature output of the tth layer, Transformer recursive is the recursive transformer module; S25. Based on the dynamic self-attention adjustment module, dynamically adjust the weight according to the importance of the recursive feature; the expression of the dynamic adjustment is as follows: Where: W adjusted is the adjusted weight feature, β j is the dynamic weight coefficient, R out，j is the j-th layer output of the recursive feature, J is the number of recursive layers, W γ To adjust the weight matrix, b γ To adjust the bias term; S26. Input the adjusted weight features into the progress feature classification module to generate progress features of the power transmission and transformation project at different stages, complete the training of the deep learning model, and obtain the deep learning model of the power transmission and transformation project; The expressions of the progress characteristics of the substation engineering at different stages are as follows: P stage (i)=Softmax(W c ·W adjusted +b c ); Where: P stage (i) is the transmission and transformation project progress feature of the i-th sample, Softmax is the activation function, W c is the classifier weight matrix, b c is the classifier bias.

8. The power transmission and transformation project progress identification and analysis system based on deep learning according to claim 7 is characterized by: The key factor identification module (4) identifies the key factors according to the following steps: S41. Compare the current progress status with the actual progress of the power transmission and transformation project to obtain a progress difference; the expression of the progress difference is as follows: ReLU(P stage (i,j)-P plan (i,j))) Where: D diff (i) is the progress difference of the i-th sample, P stage (i, j) is the jth item of the current progress status feature, P plan (i, j) is the jth item of the actual transmission and transformation project schedule characteristic, w j is the weight of the jth item, N is the total number of progress feature items, and ReLU is the rectified linear unit function; S42, analyzing the progress difference, identifying the lagging or leading part of the progress of the power transmission and transformation project; the expressions of the lagging or leading part are as follows: Where: D lag (i) is the delayed progress part, D lead (i) is the part of the progress ahead, α j and β j is the weight of the corresponding item, Sigmoid is the S-type activation function; S43. Combine the work order data and perform progress analysis on the lagging or leading parts. The expression of the progress analysis is as follows: Among them: A task (i) is the progress analysis result after combining the work ticket, T task (i, j) is the jth item of the job ticket data, γ j is the weight of the corresponding item, Tanh is the hyperbolic tangent function; S44. Combine the financial budget and expenditure data to perform progress analysis on the delayed or advanced parts. The expression of the progress analysis is as follows: LeakyReLU(D diff (i, j))) where: A finance (i) is the progress analysis result combined with financial data, F budget (i, j) is the jth item of financial budget data, F expense (i, j) is the jth item of financial expenditure data, δ j is the weight of the corresponding item, ∈ is the minimum value to prevent division by zero error, and LeakyReLU is the rectified linear unit function with leakage; S45. Based on the progress analysis results combined with the work order, financial budget and expenditure data, a comprehensive analysis is performed to obtain the key factors affecting the progress of the power transmission and transformation project; the expression of the key factors is as follows: Softmax(A task (i,j)+A finance (i,j))) Where: K factor (i) is the key factor of the i-th sample, η j is the weight of the corresponding item, and Softmax is the normalized exponential function.

9. The power transmission and transformation project progress identification and analysis system based on deep learning according to claim 7, characterized in that: The comprehensive trend analysis module (5) analyzes the comprehensive trend according to the following steps: S51. Predict the future progress of the power transmission and transformation project based on the deep learning model of the power transmission and transformation project, and weight the progress prediction results based on key factors; the weighted expression is as follows: P future (i,t)=M trained (D final (i),t); P weighted (i,t)=P future (i,t)×K factor (i); Where: P future (i, t) is the future power transmission and transformation project progress forecast result, D final is the data set of power transmission and transformation project progress, M trained For the deep learning model of power transmission and transformation engineering, P weighted (i, t) is the weighted result of the progress prediction of the i-th sample at the future time t; S52, performing progress trend analysis on the weighted result of progress forecast; the expression of the progress trend analysis is as follows: Where: T trend (i) is the progress trend analysis result of the i-th sample, P current (i) is the characteristic of the current progress status, and T is the total length of the prediction time period; S53, performing difference analysis on the video image data of the construction sites with different voltage levels; the expression of the difference analysis is as follows: Where: D compare (i, v) is the difference analysis result of the i-th sample under different voltage levels V, V data (i, k, v) is the video image data, V baseline (i, k, v) is the baseline data, K is the total number of data items; S54. Based on the progress trend analysis result and the difference analysis result, a comprehensive trend analysis is performed; the expression of the comprehensive trend analysis is as follows: T comprehensive (i)=α·T trend (i)+β·D compare (i,v); Where: T comprehensive (i) is the comprehensive trend analysis result of the i-th sample, and α and β are weight coefficients.

10. A power transmission and transformation project progress identification and analysis device based on deep learning, characterized in that: The device comprises a processor (6) and a memory (7); The memory (7) is used to store computer program code (71) and transmit the computer program code (71) to the processor (6); The processor (6) is used to execute the power transmission and transformation project progress identification and analysis method based on deep learning as described in any one of claims 1 to 5 according to the instructions in the computer program code (71).

Images