A method and device for predicting the concrete consumption of a power transmission line tower project

Abstract

The present application relates to a kind of transmission line tower engineering concrete consumption prediction method and device, belong to electric power engineering technical field, solve the problem that the spatial region characteristic of existing method is not considered concrete consumption and lead to inaccurate measurement and model precision etc..The method comprises: obtaining the concrete consumption of transmission line tower engineering and its historical influence factor data;The clustering analysis is carried out to historical influence factor data, and key influence factor set under a variety of topography and geology conditions is formed as training set in combination with correlation analysis;Deep confidence network model is constructed, and the improved grey wolf algorithm is applied to optimize the connection weight of deep confidence network model;The optimized deep confidence network model is trained using training set to obtain concrete consumption prediction model;And the real-time key influence factor data to be predicted is obtained and input into prediction model to predict concrete consumption.By considering the spatial region characteristic of influence factor, the precision of measurement and prediction model is improved.

Description

Technical Field

[0001] This invention relates to the field of power engineering technology, and in particular to a method and apparatus for predicting concrete consumption in transmission line tower engineering. Background Technology

[0002] The construction of tower foundation concrete is a crucial component of power transmission line engineering and a fundamental guarantee for the safe operation of transmission lines. In recent years, with the continuous implementation of power grid companies' strategies for improving quality and efficiency, strengthening cost control in power transmission line engineering, while ensuring construction quality, has become a key focus for these companies. Differences in production consumption are undoubtedly one of the important reasons for the differences in the cost levels of power transmission line engineering projects.

[0003] Current research primarily focuses on evaluating concrete technical parameters such as aggregate alkali content, material proportions, construction quality, health status, structural integrity and durability, and earthquake damage prevention. However, the calculation and optimization techniques for concrete consumption in building engineering, especially in transmission line tower foundations, remain largely unexplored. Furthermore, the selection of evaluation methods often involves constructing corresponding evaluation index systems, determining index weights, and employing specific evaluation methods, leading to significant human bias in the evaluation conclusions. Summary of the Invention

[0004] Based on the above analysis, the present invention aims to provide a method and apparatus for predicting concrete consumption in transmission line tower engineering, in order to solve the problems of inaccurate calculation and evaluation models caused by existing methods not considering the spatial regional characteristics of concrete consumption in transmission line foundation engineering.

[0005] On one hand, embodiments of the present invention provide a method for predicting concrete consumption in transmission line tower projects, comprising: acquiring concrete consumption data of transmission line tower projects and historical influencing factors of concrete consumption; performing cluster analysis on the historical influencing factors of concrete consumption based on various topographic and geological conditions divided by the spatial regional distribution of transmission line tower projects, and combining correlation analysis to form a set of key influencing factors of concrete consumption of transmission line projects under various topographic and geological conditions as a training set, wherein the various topographic and geological conditions include plateaus, mountains, hills, plains and deserts; constructing a deep belief network model, and applying an improved gray wolf algorithm to optimize the connection weights of the deep belief network model; training the optimized deep belief network model using the training set to obtain a concrete consumption prediction model; and acquiring real-time key influencing factor data of concrete consumption of the transmission line tower project to be predicted and inputting it into the concrete consumption prediction model to predict the concrete consumption.

[0006] The beneficial effects of the above technical solution are as follows: After collecting the influencing factors of concrete consumption in power transmission line foundation engineering, the present invention first performs cluster analysis on the influencing factors based on the FCM method, fully considering the spatial regional characteristics of the influencing factors, thereby improving the accuracy of the calculation and evaluation model.

[0007] Based on further improvements to the above method, the historical influencing factors data of concrete consumption include topography, geology, altitude, temperature, wind speed, icing, tower height, engineering foundation pit excavation depth, construction procedures, construction technology, foundation engineering voltage level, foundation steel quantity, foundation concrete quantity, and construction equipment data.

[0008] Further improvements to the above method include, before performing cluster analysis on the historical influencing factors data of concrete consumption, the following steps are taken: A particle swarm optimization algorithm based on genetics is applied to preprocess the concrete consumption data of transmission line tower projects to extract valid data and remove invalid and redundant data. Specifically, the influencing factors data of concrete consumption are classified, and an initial population is generated by equally dividing the limit ranges of the parameters. Then, the parameters are binary encoded using the following formula:

[0009]

[0010] Among them, parameter A to be corrected ri The range of variation is [A] MIN A MAX ], b is the parameter A to be corrected ri The binary number, which contains all parameters A to be corrected. r1 A r2 ...A rm The binary codes are concatenated to form a binary string p. If there are mk parameters to be identified, and the binary code of each parameter is q bits, then the binary string p has a total of (mk)q bits. A fitness function is established, where the individual fitness f is:

[0011] f = C / [D + f(e)],

[0012] Where C and D are proportionality coefficients and non-zero adjustment parameters used to adjust the fitness value, and f(e) is the error function. The smaller the value of the error function, the larger the fitness value of the individual f, and the greater the probability that the individual will be retained; conversely, the individual will be eliminated. The individual with the worst fitness in each generation is replaced by the best individual in each iteration and the existing best individual to change the convergence speed of the algorithm. During the execution of the algorithm, the best individual in the parent generation is saved and directly copied to avoid the loss of the individual with the best fitness in the parent generation during selection, crossover, and mutation. The adaptive crossover rate is calculated using the following formula:

[0013]

[0014] Among them, P c1 P c2 The value of f ranges from [0,1]. max The maximum fitness value in each generation of the population, f avg f' is the average fitness value, and f' is the larger fitness value among the individuals to be crossed.

[0015] Based on further improvements to the above method, and according to the various topographic and geological conditions categorized by the spatial distribution of transmission line tower projects, a cluster analysis is performed on the historical influencing factors data of concrete consumption. This includes: using fuzzy C-means clustering to perform cluster analysis on the influencing factors data of concrete consumption under the topographic and geological conditions of the plateau, mountains, hills, plains, and deserts; and forming a first set of influencing factors under the plateau condition, a second set of influencing factors under the mountains condition, a third set of influencing factors under the hills condition, a fourth set of influencing factors under the plains condition, and a fifth set of influencing factors under the desert condition based on the cluster analysis results.

[0016] Based on further improvements to the above method, and combined with correlation analysis, a set of key influencing factors for concrete consumption in transmission line projects under different terrain and geological conditions is formed as a training set. This set includes the following: The correlation coefficient R between each historical influencing factor value and the corresponding concrete consumption in the preprocessed historical influencing factor data is calculated using the following formula:

[0017]

[0018] Where, x i Indicates the value of the influencing factor. This represents the average value of this type of influencing factor; y i This indicates the corresponding concrete volume. The value represents the average concrete volume; i represents the selected sample project, where i = 1, 2, ..., n; based on the correlation coefficient R, key influencing factors are selected from the first set of influencing factors under plateau conditions, the second set of influencing factors under mountain conditions, the third set of influencing factors under hilly conditions, the fourth set of influencing factors under plain conditions, and the fifth set of influencing factors under desert conditions to form a training set, wherein the key influencing factors include: voltage level, geology, topography, conductor cross-section, number of splits, and tension ratio; and the training set includes: a first set of key influencing factors corresponding to the first set of influencing factors, a second set of key influencing factors corresponding to the second set of influencing factors, a third set of key influencing factors corresponding to the third set of influencing factors, a fourth set of key influencing factors corresponding to the fourth set of influencing factors, and a fifth set of key influencing factors corresponding to the fifth set of influencing factors.

[0019] Based on a further improvement of the above method, the deep belief network model includes: a first RBM layer, a second RBM layer, a third RBM layer and an output layer connected in sequence, wherein each of the first RBM layer, the second RBM layer and the third RBM layer includes a visible layer and a hidden layer.

[0020] Further improvements to the above method include training an optimized deep belief network model using the training set: inputting the first, second, third, fourth, and fifth key influencing factor sets from the training set as input layer vectors into the first RBM layer to complete unsupervised training; after the first RBM layer completes feature learning from the training set, it obtains first feature data, which is then used as the input vector for the second RBM layer to continue unsupervised training; then, it obtains second feature data from the second RBM layer and third feature data from the third RBM layer as output features, and obtains the local optimal parameters in each RBM layer; and uses the backpropagation algorithm to perform supervised fine-tuning of multiple RBM layers from back to front, adjusting the parameters of each RBM layer to obtain the global optimal parameters of the deep belief network model.

[0021] Based on the above method, further improvements to the gray wolf algorithm are made to optimize the connection weights of the deep belief network model. The improved gray wolf algorithm includes the following encirclement, hunting, and attack processes, wherein, during the encirclement process, the data model of the gray wolf encircling its prey is represented by the following formula:

[0022] D = |C·X p (t)-X(t)|;

[0023] X(t+1)=X p (t)-A·D;

[0024] Where D represents the distance between the wolf pack and the prey, A = 2α·r1 - α, C = 2·r2, t represents the number of iterations, and X p X(t) represents the position vector of the prey and the position of the wolf pack, X(t) represents the current position vector of the gray wolf, X(t+1) represents the position vector of the optimal gray wolf, r1 and r2 are random variables with values ​​ranging from [0,1], and α has values ​​ranging from [0,2]. Individual gray wolves are represented by α, β, δ, and ω, where α represents the individual making decisions and managing the wolf pack, β and δ have lower fitness than α, and ω represents ordinary individuals.

[0025] During the hunting process, the optimal positioning of individual gray wolves α, β, and δ is performed, and the approximate distance D between individual gray wolf α, β, and δ and the current position X is determined. α D β and D δ They are represented as follows:

[0026] D α =|C1·X α -X|;

[0027] D β =|C2·X β -X|;

[0028] D δ =|C3·X δ -X|;

[0029] Among them, X α X β and X δ Let C1, C2, and C3 represent the positions of the global optimal solution, the second solution, and the third solution, respectively; C1, C2, and C3 represent random vectors with values ​​ranging from [0,1], and X and X(t+1) are represented by the following formulas:

[0030] X1 = X α -A1·(D α ),

[0031] X2 = X β -A2·(D β ),

[0032] X3 = X δ -A3·(D δ ),

[0033]

[0034] Where X(t+1) represents the updated solution, and A1, A2, and A3 represent random vectors;

[0035] D α D β and D δ Let A and B represent the distances between the current candidate gray wolves and the three optimal wolves, respectively. During the attack, the attack is the final stage of the wolf pack's hunting behavior, achieved by adjusting the individual gray wolf parameter α. If A ≤ 1, the wolf pack approaches the prey and concentrates its attack on the prey (X). * ,Y * Conversely, the wolf pack gradually moves away from its prey.

[0036] Based on further improvements to the above method, a fuzzy C-means clustering algorithm is used to perform cluster analysis on the influencing factors of concrete consumption under the topographic and geological conditions of the plateau, mountains, hills, plains, and deserts. This includes assuming the sample set is D = {x1, x2, ..., x...} n} represents the data on factors influencing the concrete consumption of n samples, for D={x1,x2,…,x n Fuzzy clustering is performed to obtain c C1, C2, ..., C... c P = {P1, P2, ..., P} c A cluster represents the set of cluster centers of all subsets, where U = (u ij ) represents the membership matrix, where u represents the membership degree. ij It is used to indicate sample x i Its subset C j The subordinate relationship should satisfy:

[0037]

[0038] The objective function of the dissimilarity index in the clustering algorithm can be defined as:

[0039]

[0040] Where J represents the sum of squared distances between instance data and cluster centers; d ij The distance between the i-th data point and the j-th cluster center is represented by a distance metric function, and the distribution of cluster centers is represented by Euclidean distance d. ij =||x i -p j ||;

[0041] Construct a new objective function, and set the necessary condition for minimizing the objective function of the dissimilarity index as follows:

[0042]

[0043] λ j(j = 1, 2, ..., n) are n Lagrange multipliers of the constraint equations, m is used to determine the smoothing factor of the fuzziness level of the membership matrix U, and the fuzziness level of U is positively correlated with the value of m. The necessary condition for minimizing the objective function of the dissimilarity index with respect to all input parameters is:

[0044]

[0045]

[0046] The optimization is completed by continuously iterating until the target conditions are met. At this point, the objective function J reaches its minimum value. The cluster centers and membership degrees are continuously adjusted during the iteration process until the iteration conditions are met.

[0047] On the other hand, embodiments of the present invention provide a device for predicting concrete consumption in transmission line tower projects, comprising: a data acquisition module, used to acquire concrete consumption of transmission line tower projects and historical influencing factor data of concrete consumption, and to acquire real-time key influencing factor data of concrete consumption of the transmission line tower project to be predicted; a training set generation module, used to perform cluster analysis on the historical influencing factor data of concrete consumption according to various terrain and geological conditions divided by the spatial regional distribution of transmission line tower projects, and to combine correlation analysis to form a set of key influencing factors of concrete consumption of transmission line projects under various terrain and geological conditions as a training set, wherein the various terrain and geological conditions include plateaus, mountains, hills, plains and deserts; a network model, used to construct a deep belief network model, and apply an improved gray wolf algorithm to optimize the connection weights of the deep belief network model; a training module, used to train the optimized deep belief network model using the training set to obtain a concrete consumption prediction model; and a prediction module, used to input real-time key influencing factor data into the concrete consumption prediction model to predict concrete consumption.

[0048] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0049] 1. Existing methods do not consider the significant spatial regional characteristics of concrete consumption in transmission line foundation engineering. That is, the influencing factors on concrete consumption in tower foundation engineering vary under different external conditions such as terrain, geology, altitude, and wind speed, leading to inaccuracies in the calculation and evaluation models. Therefore, this invention, after collecting the influencing factors of concrete consumption in transmission line foundation engineering, first performs cluster analysis on these factors based on the FCM method, fully considering the spatial regional characteristics of the influencing factors, thereby improving the accuracy of the calculation and evaluation models.

[0050] 2. In the selection of the evaluation and analysis model, considering the problems of the model itself, the Grey Wolf algorithm was used to optimize the design of the model parameters, so that it is more in line with the characteristics of concrete consumption measurement and analysis of transmission line foundation engineering.

[0051] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0052] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0053] Figure 1 This is a flowchart illustrating the method for predicting concrete consumption in transmission line tower engineering according to an embodiment of the present invention.

[0054] Figure 2 A flowchart of a method for predicting concrete consumption in transmission line tower engineering according to an embodiment of the present invention;

[0055] Figure 3 This is a network topology diagram of a deep belief network model (DBN) according to an embodiment of the present invention;

[0056] Figure 4 This is a block diagram of a device for predicting concrete consumption in transmission line tower engineering according to an embodiment of the present invention. Detailed Implementation

[0057] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0058] A specific embodiment of the present invention discloses a method for predicting the concrete consumption of transmission line tower projects. For example... Figure 2As shown, the method for predicting concrete consumption in transmission line tower projects includes: in step S202, acquiring concrete consumption data and historical influencing factors data for transmission line tower projects; in step S204, performing cluster analysis on historical influencing factors data for concrete consumption based on various topographic and geological conditions categorized by the spatial distribution of transmission line tower projects, and combining correlation analysis to form a set of key influencing factors for concrete consumption under various topographic and geological conditions as a training set, wherein the various topographic and geological conditions include plateaus, mountains, hills, plains, and deserts; in step S206, constructing a deep belief network model and applying an improved gray wolf algorithm to optimize the connection weights of the deep belief network model; in step S208, using the training set to train the optimized deep belief network model to obtain a concrete consumption prediction model; and in step S210, acquiring real-time key influencing factor data for concrete consumption of the transmission line tower project to be predicted and inputting it into the concrete consumption prediction model to predict the concrete consumption.

[0059] Compared with the prior art, the method for predicting concrete consumption in transmission line tower engineering provided in this embodiment first performs cluster analysis on the influencing factors of concrete consumption in transmission line foundation engineering based on the FCM method after collecting the influencing factors, and fully considers the spatial regional characteristics of the influencing factors, thereby improving the accuracy of the calculation and evaluation model.

[0060] In the following text, refer to Figure 2 and Figure 3 The following describes in detail each step of the method for predicting concrete consumption in transmission line tower engineering according to an embodiment of the present invention.

[0061] In step S202, the concrete consumption of the transmission line tower project and the historical influencing factors of concrete consumption are obtained. Specifically, the historical influencing factors of concrete consumption include topography, geology, altitude, temperature, wind speed, icing, tower height, excavation depth of the foundation pit, construction procedures, construction technology, voltage level of the foundation project, amount of foundation steel, amount of foundation concrete, and construction equipment data.

[0062] Then, before performing cluster analysis on the historical influencing factors data of concrete consumption, the following steps are included: applying a particle swarm optimization algorithm based on genetics to preprocess the concrete consumption data of transmission line tower projects to extract effective data and remove invalid and redundant data. Specifically, the influencing factors data of concrete consumption are classified, and the limit ranges of parameters are equally divided to generate an initial population. Then, the parameters are binary encoded using the following formula:

[0063]

[0064] Among them, parameter A to be corrected ri The range of variation is [A] MIN A MAX ], b is the parameter A to be corrected ri The binary number, which contains all parameters A to be corrected. r1 A r2 ...A rm The binary codes are concatenated to form a binary string p. If there are mk parameters to be identified, and the binary code of each parameter is q bits, then the binary string p has a total of (mk)q bits.

[0065] Establish a fitness function, where the individual fitness f is:

[0066] f = C / [D + f(e)],

[0067] Wherein, C and D are proportionality coefficients and non-zero adjustment parameters used to adjust the fitness value, and f(e) is an error function. The smaller the value of the error function, the larger the individual fitness value f, and the greater the probability that the individual will be retained. Conversely, the individual will be eliminated.

[0068] The worst-fit individual in each generation is replaced with the best individual from each iteration and the existing best individual to change the convergence speed of the algorithm. During the algorithm execution, the best individual from the parent generation is saved and directly copied to avoid the loss of the best-fit individual from the parent generation during selection, crossover, and mutation. The adaptive crossover rate is calculated using the following formula:

[0069]

[0070] Among them, P c1 P c2 The value can be set manually; specifically, P c1 P c2 The value is determined based on importance, and the range is [0,1]. max The maximum fitness value in each generation of the population, f avg f' is the average fitness value, and f' is the larger fitness value among the individuals to be crossed.

[0071] In step S204, based on the spatial distribution of transmission line tower projects and various topographic and geological conditions, cluster analysis is performed on the historical influencing factors data of concrete consumption. Furthermore, correlation analysis is combined to form a training set of key influencing factors for concrete consumption under various topographic and geological conditions, including plateaus, mountains, hills, plains, and deserts. Specifically, the cluster analysis of historical influencing factors data of concrete consumption based on the spatial distribution of transmission line tower projects includes: using fuzzy C-means clustering to cluster the influencing factors data of concrete consumption under the topographic and geological conditions of plateaus, mountains, hills, plains, and deserts; and forming a first influencing factor set under plateau conditions, a second influencing factor set under mountain conditions, a third influencing factor set under hill conditions, a fourth influencing factor set under plain conditions, and a fifth influencing factor set under desert conditions based on the cluster analysis results.

[0072] Fuzzy C-means clustering algorithm was used to perform cluster analysis on the influencing factors of concrete consumption under topographic and geological conditions of plateau, mountain, hill, plain and desert. The analysis included: assuming the sample set is D = {x1, x2, ..., x...} n} represents the data on factors influencing the concrete consumption of n samples, for D={x1,x2,…,x n Fuzzy clustering is performed to obtain c C1, C2, ..., C... c P = {P1, P2, ..., P} c A cluster represents the set of cluster centers of all subsets, where U = (u ij ) represents the membership matrix, where u represents the membership degree. ij It is used to indicate sample x i Its subset C j The subordinate relationship should satisfy:

[0073]

[0074] The objective function of the dissimilarity index in clustering algorithms can be defined as:

[0075]

[0076] Where J represents the sum of squared distances between instance data and cluster centers; d ij The distance between the i-th data point and the j-th cluster center is represented by a distance metric function, and the distribution of cluster centers is represented by Euclidean distance d. ij =||x i -p j ||;Construct a new objective function as follows, which is a necessary condition for minimizing the objective function of the dissimilarity index:

[0077]

[0078] Where, λ j (j = 1, 2, ..., n) are n Lagrange multipliers of the constraint equations. m is used to determine the smoothing factor of the fuzziness level of the membership matrix U. The fuzziness level of U is positively correlated with the value of m. Taking the derivative with respect to all input parameters, the necessary condition for minimizing the objective function of the dissimilarity index is:

[0079]

[0080]

[0081] The optimization is completed by continuously iterating until the target conditions are met. At this point, the objective function J reaches its minimum value. The cluster centers and membership degrees are continuously adjusted during the iteration process until the iteration conditions are met.

[0082] Specifically, a set of key influencing factors on concrete consumption in transmission line projects under different terrain and geological conditions was formed by combining correlation analysis as a training set. This set included the following: The correlation coefficient R between each historical influencing factor value and the corresponding concrete consumption in the preprocessed historical influencing factor data was calculated using the following formula:

[0083]

[0084] Where, x i Indicates the value of the influencing factor. This represents the average value of this type of influencing factor; y i This indicates the corresponding concrete volume. The value represents the average concrete volume; i represents the selected sample project, where i = 1, 2, ..., n; based on the correlation coefficient R, key influencing factors are selected from the first set of influencing factors under plateau conditions, the second set of influencing factors under mountain conditions, the third set of influencing factors under hilly conditions, the fourth set of influencing factors under plain conditions, and the fifth set of influencing factors under desert conditions to form a training set. The key influencing factors include: voltage level, geology, topography, conductor cross-section, number of splits, and tension ratio; and the training set includes: the first key influencing factor set corresponding to the first set of influencing factors, the second key influencing factor set corresponding to the second set of influencing factors, the third key influencing factor set corresponding to the third set of influencing factors, the fourth key influencing factor set corresponding to the fourth set of influencing factors, and the fifth key influencing factor set corresponding to the fifth set of influencing factors.

[0085] In step S206, a deep belief network model is constructed, and the improved Grey Wolf algorithm is applied to optimize the connection weights of the deep belief network model. Specifically, refer to... Figure 3The deep belief network model includes a first RBM layer, a second RBM layer, a third RBM layer, and an output layer connected in sequence, wherein each of the first RBM layer, the second RBM layer, and the third RBM layer includes a visible layer and a hidden layer.

[0086] The improved Grey Wolf algorithm is used to optimize the connection weights of a deep belief network model. The improved Grey Wolf algorithm includes the following encirclement, hunting, and attack processes. In the encirclement process, the data model of the Grey Wolf encircling its prey is represented by the following formula:

[0087] D = |C·X p (t)-X(t)|,

[0088] X(t+1)=X p (t)-A·D,

[0089] Where D represents the distance between the wolf pack and the prey, A = 2α·r1 - α, C = 2·r2, t represents the number of iterations, and X p (t) represents the position vector of the prey and the position of the wolf pack, i.e., X p Let X(t) represent the current position vector of the gray wolf, and X(t+1) represent the optimal position vector of the gray wolf. r1 and r2 are random variables with values ​​ranging from [0,1], and α has a value range of [0,2]. Let α, β, δ, and ω represent individual gray wolves, where α represents the individual making decisions and managing the wolf pack, β and δ have lower fitness than α, and ω represents ordinary individuals. During the hunting process, the gray wolf individuals α, β, and δ are optimized for positioning. The approximate distance D between the gray wolf individuals α, β, and δ and their current position X is calculated. α D β and D δ They are represented as follows:

[0090] D α =|C1·X α -X|,

[0091] D β =|C2·X β -X|,

[0092] D δ =|C3·X δ -X|,

[0093] Among them, X α X β and X δ Let C1, C2, and C3 represent the positions of the global optimal solution, the second solution, and the third solution, respectively; C1, C2, and C3 represent random vectors with values ​​ranging from [0,1], and X and X(t+1) are represented by the following formulas:

[0094] X1 = X α-A1·(D α ),

[0095] X2 = X β -A2·(D β ),

[0096] X3 = X δ -A3·(D δ ),

[0097]

[0098] Where X(t+1) represents the updated solution, and A1, A2, and A3 represent random vectors; D α D β and D δ Let A and B represent the distances between the current candidate gray wolves and the three optimal wolves, respectively. During the attack, the attack is the final stage of the wolf pack's hunting behavior, achieved by adjusting the individual gray wolf parameter α. If A ≤ 1, the wolf pack approaches the prey and concentrates its attack on the prey (X). * ,Y * Conversely, the wolf pack gradually moves away from its prey.

[0099] In step S208, an optimized deep belief network model is trained using the training set to obtain a concrete consumption prediction model. Specifically, training the optimized deep belief network model using the training set includes: inputting the first, second, third, fourth, and fifth key influencing factor sets from the training set as input layer vectors into the first RBM layer to complete unsupervised training; after the first RBM layer completes feature learning from the training set, it obtains the first feature data, which is then used as the input vector for the second RBM layer to continue unsupervised training; then, the second feature data is obtained from the second RBM layer, and the third feature data is obtained from the third RBM layer as output features, and the local optimal parameters in each RBM layer are obtained; the error backpropagation algorithm is used to perform supervised fine-tuning of multiple RBM layers from back to front, adjusting the parameters of each RBM layer to obtain the globally optimal parameters of the deep belief network model.

[0100] In step S210, real-time key influencing factor data of concrete consumption for the transmission line tower project to be predicted are obtained and input into the concrete consumption prediction model to predict the concrete consumption.

[0101] Another specific embodiment of the present invention discloses a device for predicting the concrete consumption of transmission line tower projects. (Reference) Figure 4The device for predicting concrete consumption in transmission line tower projects includes: a data acquisition module 402, used to acquire concrete consumption data and historical influencing factors data of concrete consumption in transmission line tower projects, and to acquire real-time key influencing factor data of concrete consumption in transmission line tower projects to be predicted; a training set generation module 404, used to perform cluster analysis on historical influencing factor data of concrete consumption based on various terrain and geological conditions divided by the spatial distribution of transmission line tower projects, and to combine correlation analysis to form a set of key influencing factors of concrete consumption in transmission line projects under various terrain and geological conditions as a training set, wherein various terrain and geological conditions include plateaus, mountains, hills, plains and deserts; a network model 406, used to construct a deep belief network model, and apply an improved gray wolf algorithm to optimize the connection weights of the deep belief network model; a training module 408, used to train the optimized deep belief network model using the training set to obtain a concrete consumption prediction model; and a prediction module 410, used to input real-time key influencing factor data into the concrete consumption prediction model to predict concrete consumption.

[0102] The following text will refer to Figure 2 and Figure 3 The present invention provides a detailed description of a method for predicting concrete consumption in transmission line tower engineering according to specific examples.

[0103] This invention aims to provide a method for optimizing concrete consumption in transmission line tower engineering based on FCM-GWO-DBN, in order to solve the problem of optimizing the accuracy of concrete consumption in transmission line foundation engineering by analyzing the main influencing factors of concrete consumption under different terrain and geological conditions.

[0104] refer to Figure 1 The method for optimizing concrete consumption in transmission line tower foundation engineering based on FCM-GWO-DBN includes the following steps.

[0105] In step S102, the data on factors affecting the amount of concrete consumed in the foundation engineering of the transmission line tower are determined, including: topography and geology, altitude, temperature, wind speed, icing, tower height, foundation pit excavation depth, construction procedures, construction technology, and construction equipment data.

[0106] In step S104, the NACEMD method is applied to preprocess the data and remove outlier data. In an optional embodiment, a hybrid optimization algorithm based on particle swarm optimization and genetics is applied to process the concrete consumption data of transmission line tower projects, extract valid data, and remove invalid and redundant data, which can not only improve data processing efficiency but also improve prediction accuracy.

[0107] In step S106, considering the significant spatial regional characteristics of concrete consumption in transmission line foundation engineering, the spatial distribution of transmission line projects is first divided into five terrain conditions: plateau, mountain, hill, plain, and desert. Then, the Functional Cluster Method (FCM) is applied to cluster the influencing factor data to determine the set of key influencing factors. Specifically, the FCM method is used to cluster the influencing factor data under different terrain and geological conditions, combined with correlation analysis, to form a set of key influencing factors for concrete consumption in transmission line foundation engineering under different terrain and geological conditions. This results in cluster analysis of transmission line projects under different terrain and geological conditions, enabling grouping based on the similarity of concrete consumption per unit foundation, providing a foundation for subsequent prediction model construction and improving prediction accuracy.

[0108] For example, the fuzzy C-means clustering method (FCM) was applied to perform cluster analysis on the influencing factor data under five different topographic and geological conditions: plateau, hills, mountains, plains, and desert, to determine the set of influencing factors under the main different topographic and geological conditions. Five indicators—voltage level of the foundation engineering, topography, geology, amount of foundation steel, and amount of foundation concrete—were extracted as feature quantities for fuzzy clustering.

[0109] Fuzzy C-means clustering (FCM) is an algorithm that uses membership functions to determine the degree to which a sample point belongs to a certain class. The main principle of FCM is as follows:

[0110] Suppose the sample set is D = {x1, x2, ..., x...} n} represents the feature vector set of n samples. Fuzzy clustering of D yields c subsets C1, C2, ..., C... c P = (P1, P2, ..., P c A cluster represents the set of cluster centers of all subsets, where U = (u ij ) represents the membership matrix, where u represents the membership degree. ij It is used to indicate sample X i Its subset C j The subordinate relationship should satisfy:

[0111]

[0112] The FCM clustering algorithm performs fuzzy clustering, and its objective function for the dissimilarity index can be defined as:

[0113]

[0114] Where J represents the sum of squared distances between instance data and cluster centers; d ijThe distance between the i-th data point and the j-th cluster center is represented by a distance metric function. The type of function chosen may differ depending on the distribution of the cluster centers; generally, Euclidean distance d is used. ij =||x i -p j ||.

[0115] Construct a new objective function as follows, which is a necessary condition for the above equation (2) to reach its minimum value:

[0116]

[0117] Here λ j (j=1,2,...,n) are the Lagrange multipliers of the n constraints in the above formula (1). m: the coefficient used to determine the fuzziness level of the membership matrix U, usually called the smoothing factor. The fuzziness level of U is positively correlated with the value of m. The necessary condition for taking the derivative of all input parameters to minimize formula (3) is:

[0118]

[0119]

[0120] The optimization is completed by iteratively calculating until the objective condition is met. At this point, the objective function J reaches its minimum value.

[0121] During the iteration process, the cluster centers and membership degrees are continuously adjusted until the iteration conditions are met. The membership degree of the FCM clustering algorithm is not rigidly defined as belonging to 1 or 0, but can take fuzzy values ​​between different categories.

[0122] Key influencing factors were identified by calculating the correlation coefficient between influencing factors and concrete consumption. Several sets of sample data were collected, outliers were processed, and dimensionless data was standardized to the range [0, 1]. Influencing factors were then eliminated using the correlation coefficient R between each independent and dependent variable, removing negligible factors with little impact on the model analysis to identify key influencing factors. The formula for calculating the correlation coefficient R is as follows:

[0123]

[0124] Where, x i Indicates the value of the influencing factor. This represents the average value of this type of influencing factor; y i Indicates the output vector value. This represents the average value of the output vector. The key influencing factors selected ultimately included voltage level, geology, topography, conductor cross-section, number of splits, and tension ratio.

[0125] In step S108, a deep belief network (DBN) model is constructed, and the Grey Wolf algorithm (GWO) is applied to optimize the core parameters. Several sets of preprocessed sample data are combined into a training set to determine the optimal concrete consumption. Using the improved Grey Wolf algorithm to optimize the DBN network connection weights improves the network's classification performance and provides a better fit.

[0126] Deep learning models, as derivatives of multi-layer neural network training, differ in that they abstract low-level features of the model data, uncovering the inherent distribution characteristics of the data and using fewer training data samples to obtain the essential features of the data. At the same time, deep learning models inherit the robustness of neural network training models, while also possessing the ability to handle complex functions even with limited data samples.

[0127] Deep learning models primarily integrate Reactive Particle Swarm Optimization (RBM) and adaptive intelligent algorithms. Their training approach is as follows: 1. Extract the low-level data features of the deep learning model to serve as input variables for the top-level learning of the model, while simultaneously employing a layer-by-layer training mode from the bottom to the top levels; 2. After training to the top level of the model, use an adaptive particle swarm optimization algorithm to adaptively optimize and adjust the entire training network, ensuring that the model training results can escape local solutions.

[0128] A concrete consumption analysis model based on a deep belief network (DBN) is constructed. The DBN network topology is as follows: Figure 3 As shown, each RBM layer consists of a visible layer v i and hidden layer h i The network consists of several layers. The last layer is a BP network, which is mainly used for fine-tuning the weights from top to bottom.

[0129] In the RBM, given (v,h), the energy function is:

[0130]

[0131] In the above formula, θ = {w, a, b} are network parameters, w is the weight between the visible layer and the hidden layer, a and b are the biases between the visible layer and the hidden layer, and m and n are the number of neurons in the visible layer and the hidden layer, respectively.

[0132] Based on the energy function, the following joint probability distribution function can be obtained:

[0133]

[0134] In the above formula, It is represented as the normalization factor, which is the algebraic sum of the energy functions of all variables.

[0135] Once the v-state of the visible layer is determined, the activation probability of the hidden layer units is:

[0136]

[0137] Once the h-state of the hidden layer is determined, the activation probability of the visible layer cells is:

[0138]

[0139] When the number of training samples is K, the parameter θ can be determined by solving the log-likelihood function maximization problem. The objective function of the log-likelihood function maximization problem is given by equation (11):

[0140]

[0141] In the above formula, maxL(θ) is obtained by the stochastic gradient method.

[0142] By repeating the Gibbs sampling, the update rule for the RBM parameters can be obtained as shown in equation (12):

[0143]

[0144] In the formula, ε is the RBM learning rate, <·> data and <·> recon These represent the mathematical expectations of the input data and the reconstructed data, respectively.

[0145] The DBN model training steps are as follows:

[0146] (1) Input the original data as the input layer vector into the first layer RBM to complete unsupervised training;

[0147] (2) When the first layer of RBM completes feature learning of the original data, it will obtain feature data, which will be used as the input vector of the new layer and input into the next layer of RBM to continue unsupervised training.

[0148] (3) Repeat steps (1) and (2) until each layer of RBM is trained and learned. Then the features obtained in the last layer of RBM can be used as output features, and the local optimal parameters in each layer of RBM can be obtained.

[0149] (4) The backpropagation algorithm is used to perform supervised fine-tuning of the RBM from top to bottom, adjusting the parameters of each RBM layer, and finally obtaining the global optimal parameters of the entire DBN network model. The DBN network uses several RBM units to form a basic network architecture, so that the DBN network has both unsupervised pre-training and supervised fine-tuning. This combination of unsupervised and supervised training not only solves the gradient vanishing problem in traditional methods, but also solves the problem that the network is prone to getting trapped in local optima.

[0150] The Grey Wolf Algorithm (GWO) is applied to optimize the DBN model. The specific principle of the Grey Wolf Algorithm is as follows:

[0151] In the standard GWO algorithm, gray wolf individuals are represented by α, β, δ, and ω, where α represents the individual making decisions and managing the pack, β and δ have lower fitness than α, and ω represents ordinary individuals. Specific behaviors in the GWO algorithm include encirclement, hunting, and attack.

[0152] (1) Encirclement behavior

[0153] The data model of a gray wolf surrounding its prey can be represented by equations (13) and (14).

[0154] D = |C·X p (t)-X(t)| (13)

[0155] X(t+1)=X p (t)-A·D (14)

[0156] In the formula, D represents the distance between the wolf pack and the prey, A = 2α·r1 - α, C = 2·r2, t represents the number of iterations, and X p X and r1 represent the positions of the prey and the wolf pack, respectively. r1 and r2 are random variables with a value range of [0,1], and α has a value range of [0,2].

[0157] (2) Hunting behavior

[0158] Assuming α, β, and δ represent the global optimal solution, the second solution, and the third solution of an individual gray wolf, the distances for optimizing their location are expressed as Equations (15), (16), and (17), respectively.

[0159] D α =|C1·X α -X| (15)

[0160] D β =|C2·X β -X| (16)

[0161] D δ =|C3·X δ -X| (17)

[0162] In the formula, D α D β D δ This represents the approximate distance between individuals α, β, and δ and their current position X, where X is the approximate distance between them. α X β X δThe positions of the global optimal solution, the second solution, and the third solution are represented in sequence; C1, C2, and C3 represent random vectors with values ​​ranging from [0,1]. X and X(t+1) represent equations (18), (19), (20), and (21), respectively.

[0163] X1 = X α -A1·(D α (18)

[0164] X2 = X β -A2·(D β (19)

[0165] X3 = X δ -A3·(D δ (20)

[0166]

[0167] In the formula, X(t+1) represents the updated solution, and A1, A2, and A3 represent random variables.

[0168] (3) Aggressive behavior

[0169] The attack is the final stage of the wolf pack's hunting behavior, and it can be achieved by adjusting the parameter α.

[0170] If A≤1, the wolf pack approaches the prey and concentrates its attack on the prey (X). * ,Y * Conversely, the wolf pack gradually moves away from its prey.

[0171] The standard GWO algorithm typically initializes the population positions randomly. This method may lead to an excessively large search range for the wolf pack, resulting in a longer search time. This patent generates its counterpart individuals simultaneously with the randomly generated initial population, and compares the fitness of the counterpart individual with that of the original individual. If the fitness of the counterpart individual is better than that of the original individual, the counterpart individual is used; otherwise, the original individual is used. The expression for the counterpart individual's position vector is as follows.

[0172] X'=L b +U b -X

[0173] In the formula: X′ represents the position vector of the opposing individual; X represents the position vector of the original individual; L b U b Let X be the upper and lower bounds.

[0174] As can be seen from the original formula, parameter A is determined by α, which decreases linearly from 2 to 0. However, in the actual algorithm iteration process, the search space is large in the early stage of the iteration, requiring a global search, so the rate of α decrease should be slowed down. In the later stage of the iteration, the rate of α decrease should be accelerated to improve local optimization and speed up convergence.

[0175] The convergence factor update formula based on the cosine variation law is as follows:

[0176]

[0177] In the formula: α max The maximum convergence factor is usually 2; t max n is the maximum number of iterations; n is the decreasing exponent, 0 <n<1。

[0178] In step S110, the design scheme is improved based on the results of determining the amount of concrete consumed.

[0179] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0180] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting concrete consumption in transmission line tower engineering, characterized in that, include: Obtain data on the concrete consumption of transmission line tower projects and the historical influencing factors of the concrete consumption. Based on the spatial distribution of transmission line tower projects and various topographic and geological conditions, cluster analysis was performed on the historical influencing factors of concrete consumption, and correlation analysis was combined to form a set of key influencing factors of concrete consumption in transmission line projects under various topographic and geological conditions as a training set. The various topographic and geological conditions include plateaus, mountains, hills, plains and deserts. A deep belief network model is constructed, and the connection weights of the deep belief network model are optimized by applying the improved Grey Wolf algorithm. The optimized deep belief network model is trained using the training set to obtain a concrete consumption prediction model; and The real-time key influencing factor data of the concrete consumption of the transmission line tower project to be predicted are obtained and input into the concrete consumption prediction model to predict the concrete consumption.

2. The method for predicting concrete consumption in transmission line tower engineering according to claim 1, characterized in that, The historical data on factors influencing concrete consumption include topography, geology, altitude, temperature, wind speed, icing, tower height, excavation depth of foundation pit, construction procedures, construction technology, voltage level of foundation engineering, amount of foundation steel, amount of foundation concrete, and construction equipment data.

3. The method for predicting concrete consumption in transmission line tower engineering according to claim 2, characterized in that, Before performing cluster analysis on the historical influencing factors of concrete consumption, the process also includes: applying a particle swarm optimization algorithm based on genetics to preprocess the concrete consumption data of transmission line tower projects to extract valid data and remove invalid and redundant data. The data on factors affecting concrete consumption are categorized, and an initial population is generated by equally dividing the limit ranges of the parameters. Then, the parameters are binary encoded using the following formula: ， Among them, the parameters to be corrected A ri The range of variation is [ A MIN , A MAX ], b The parameter to be corrected A ri The binary number, containing all parameters to be corrected. A r1 、A r2 … A rm The binary codes are concatenated to form a binary string p. If there are mk parameters to be identified, and the binary code of each parameter is q bits, then the binary string p has a total of (mk)q bits. Establish a fitness function, where individual fitness is... f for: f = C / [ D + f ( e )]， in, C , D These are proportional coefficients and non-zero adjustment parameters used to adjust the fitness value; f ( e Let be the error function, and the smaller the value of the error function, the higher the individual fitness. f The larger the value, the greater the probability that the individual will be retained; conversely, the individual will be eliminated. The worst-fit individual in each generation is replaced with the best individual from each iteration and the existing best individual to change the convergence speed of the algorithm. During the algorithm execution, the best individual from the parent generation is saved and directly copied to avoid the loss of the best-fit individual from the parent generation during selection, crossover, and mutation. The adaptive crossover rate is calculated using the following formula: ， in, P c1 , P c2 The value ranges from [0, 1]. f max The maximum fitness value in each generation of the population. f avg It is the average fitness value. f ' is the larger value of fitness among the individuals to be crossed.

4. The method for predicting concrete consumption in transmission line tower engineering according to claim 3, characterized in that, Based on the spatial distribution of transmission line tower projects and various topographic and geological conditions, cluster analysis was performed on the historical influencing factors of concrete consumption, including: Fuzzy C-means clustering was used to perform cluster analysis on the influencing factors of concrete consumption under the topographic and geological conditions of the plateau, mountains, hills, plains, and deserts; and Based on the cluster analysis results, the following sets of influencing factors are formed: the first set of influencing factors under plateau conditions, the second set of influencing factors under mountain conditions, the third set of influencing factors under hilly conditions, the fourth set of influencing factors under plain conditions, and the fifth set of influencing factors under desert conditions.

5. The method for predicting concrete consumption in transmission line tower engineering according to claim 4, characterized in that, Based on correlation analysis, a set of key influencing factors on concrete consumption in power transmission line projects under different terrain and geological conditions was formed as the training set, including: The correlation coefficient R between each historical influencing factor value and the corresponding concrete consumption in the pre-processed historical influencing factor data is calculated using the following formula: ， in, x i Indicates the value of the influencing factor. This represents the average value of the influencing factors; y i This indicates the corresponding concrete volume. This represents the average amount of concrete. i For the selected sample project, where, i =1, 2, ..., n ; Based on the correlation coefficient R, key influencing factors are selected from the first set of influencing factors under plateau conditions, the second set of influencing factors under mountain conditions, the third set of influencing factors under hilly conditions, the fourth set of influencing factors under plain conditions, and the fifth set of influencing factors under desert conditions to form a training set. The key influencing factors include: voltage level, geology, topography, conductor cross-section, number of splits, and tension ratio; and The training set includes: a first key influencing factor set corresponding to the first influencing factor set, a second key influencing factor set corresponding to the second influencing factor set, a third key influencing factor set corresponding to the third influencing factor set, a fourth key influencing factor set corresponding to the fourth influencing factor set, and a fifth key influencing factor set corresponding to the fifth influencing factor set.

6. The method for predicting concrete consumption in transmission line tower engineering according to claim 5, characterized in that, The deep belief network model includes: a first RBM layer, a second RBM layer, a third RBM layer, and an output layer connected in sequence, wherein each of the first RBM layer, the second RBM layer, and the third RBM layer includes a visible layer and a hidden layer.

7. The method for predicting concrete consumption in transmission line tower engineering according to claim 6, characterized in that, The optimized deep belief network model trained using the training set includes: The first set of key influencing factors, the second set of key influencing factors, the third set of key influencing factors, the fourth set of key influencing factors, and the fifth set of key influencing factors in the training set are respectively used as input layer vectors to input the first RBM layer to complete unsupervised training; After the first RBM layer completes feature learning on the training set, it obtains the first feature data. The first feature data is then used as the input vector of the second RBM layer and input into the second RBM layer to continue unsupervised training. Then, the second feature data is obtained from the second RBM layer, and the third feature data is obtained from the third RBM layer as the output feature, and the local optimal parameters in each RBM layer are obtained; The backpropagation algorithm is used to perform supervised fine-tuning of multiple RBM layers from back to front, adjusting the parameters of each RBM layer to obtain the globally optimal parameters of the deep belief network model.

8. The method for predicting concrete consumption in transmission line tower engineering according to claim 1, characterized in that, Optimizing the connection weights of the deep belief network model using the improved Grey Wolf algorithm includes the following encirclement, hunting, and attack processes, wherein... During the encirclement process, the data model of a gray wolf surrounding its prey can be represented by the following formula: D=|C•X p ( t )- X ( t ) | ； X ( t +1)= X p ( t )- A • D ； in, D Indicates the distance between the wolf pack and its prey. A =2 α • r 1- α , C =2• r 2, t Indicates the number of iterations. X p ( t () represents the position vector of the prey and the position of the wolf pack. X ( t ) represents the current position vector of the gray wolf. X ( t +1) represents the optimal position vector of the gray wolf. r 1. r 2 is a random variable whose value range is [0,1]. α The value range of is [0,2]; α , β , δ and ω Represents individual gray wolves, among which α The individual representing decision-making and management of the wolf pack. β and δ fitness is lower than α , ω For ordinary individuals: During the hunting process, individual gray wolves α , β and δ Optimize positioning, individual gray wolves α , β and δ With current location X Approximate distance D α 、D β and D δ They are represented as follows: D α = |C 1 •X α - X| ； D β = |C 2 •X β - X| ； D δ = |C 3 •X δ - X| ； in, X α 、X β and X δ The positions of the global optimal solution, the second solution, and the third solution are represented in that order. C 1. C 2. C 3 represents a random vector with values ​​ranging from [0,1]. X and are expressed by the following formula. X ( t +1): X 1= X α - A 1•( D α )， X 2= X β - A 2•( D β )， X 3= X δ - A 3•( D δ )， ， in, X ( t +1) indicates an updated solution. A 1. A 2. A 3 represents a random vector; D α 、D β and D δ These represent the distances between the current candidate gray wolves and the three optimal wolves, respectively. During the attack, which is the final stage of the wolf pack's predation behavior, individual gray wolf parameters are adjusted. α To launch an attack; if A≤1, the wolf pack approaches the prey and concentrates its attack on the prey (X). * ,Y * Conversely, the wolf pack gradually moves away from its prey.

9. The method for predicting concrete consumption in transmission line tower engineering according to claim 1, characterized in that, The fuzzy C-means clustering algorithm was used to perform cluster analysis on the influencing factors of concrete consumption under the topographic and geological conditions of the plateau, mountains, hills, plains, and deserts, including: Assume the sample set is D ={ x 1, x 2,…, x n } represents n Data on factors influencing concrete consumption for each sample, D ={ x 1, x 2,…, x n } Perform fuzzy clustering to obtain c indivual C 1. C 2、…、 C c , P ={ P 1, P 2,…, P c A cluster represents the set of cluster centers of all subsets, where U = ( u ij ) represents the membership matrix, membership degree u ij It is used to indicate the sample x i its subsets C j The subordinate relationship should satisfy: The objective function of the dissimilarity index in the clustering algorithm can be defined as: in, J The sum of squared distances between the instance data and the cluster centers; d ij The representative of the first i The data point and the first j Cluster centers are a distance metric function, and the distribution of cluster centers is determined using Euclidean distance. ; Construct a new objective function, and set the necessary condition for minimizing the objective function of the dissimilarity index as follows: λ j ( j =1, 2, ..., n ),yes A constrained Lagrange multiplier, m Used to determine the membership matrix U Smoothing factor for fuzziness level U fuzziness level and m The values ​​are positively correlated. The necessary condition for minimizing the objective function of the dissimilarity index, taking the derivative with respect to all input parameters, is: The calculation continues iteratively until the objective condition is met, at which point the objective function... Find the minimum value to complete the optimization; continuously adjust the cluster centers and membership degrees during the iteration process until the iteration conditions are met.

10. A device for predicting concrete consumption in transmission line tower engineering, characterized in that, include: The data acquisition module is used to acquire the concrete consumption of transmission line tower projects and the historical influencing factors of the concrete consumption, and to acquire the real-time key influencing factors of the concrete consumption of the transmission line tower projects to be predicted. The training set generation module is used to perform cluster analysis on the historical influencing factors of concrete consumption based on the spatial distribution of transmission line tower projects and various topographic and geological conditions, and to combine correlation analysis to form a set of key influencing factors of concrete consumption in transmission line projects under various topographic and geological conditions as the training set. The various topographic and geological conditions include plateaus, mountains, hills, plains and deserts. A network model is used to construct a deep belief network model, and the connection weights of the deep belief network model are optimized by applying an improved Grey Wolf algorithm. A training module is used to train an optimized deep belief network model using the training set to obtain a concrete consumption prediction model; and The prediction module is used to input real-time key influencing factor data into the concrete consumption prediction model to predict the concrete consumption.

Images