Equipment type selection method and system based on life cycle cost portrait system design

Abstract

The invention relates to an equipment type selection method and system based on full-life-cycle cost portrait system design. The method comprises the following steps: constructing an equipment full-life-cycle portrait label system; according to the equipment full-life-cycle portrait label system, collecting corresponding equipment full-life-cycle cost data and reliability data, and carrying out abnormal value detection and missing value interpolation through a clustering algorithm and an adversarial neural network; according to the equipment reliability label system, constructing an equipment reliability scoring system, and calculating an equipment reliability score; according to the reliability scoring result of the equipment, carrying out portrait classification on the equipment through a fuzzy C-means clustering method according to the whole life cycle portrait system of the equipment; based on portrait classification results, respectively constructing equipment full-life-cycle cost prediction models to obtain equipment full-life-cycle prediction cost; and calculating the comprehensive score of the equipment through a weighted scoring method according to the reliability scoring result of the equipment and the whole life cycle prediction cost of the equipment.

Description

Technical Field

[0001] The present application relates to the technical field of equipment selection, and more specifically, to an equipment selection method and system based on the design of a full life cycle cost portrait system. Background Art

[0002] With the growth of global energy demand and the complexity of power grid systems, equipment selection and asset management play an increasingly important role in the power industry. Traditional methods of selecting power grid equipment focus mainly on the initial investment cost, and often ignore other costs of the equipment during its life cycle, such as operation and maintenance costs, fault repair costs, and decommissioning costs. This practice may lead to unnecessary high costs for equipment during the operation phase, affecting the economic benefits and long-term reliability of the power grid system. For example, although some equipment has a low initial cost, frequent repairs and high maintenance costs may make its total life cycle cost much higher than expected, and even affect the stability of the system. On the other hand, high-reliability equipment usually means a higher initial investment, but its low maintenance cost and high stability over a long period of time can also significantly improve the overall efficiency and safety of the power grid. Therefore, how to comprehensively consider the full life cycle cost, reliability and performance of the equipment has become a key issue in the selection of power grid equipment.

[0003] In recent years, with the development of big data and artificial intelligence technologies, the selection of power grid equipment has gradually developed in the direction of data-driven and intelligent. Using big data analysis, power companies can deeply mine data such as the operating status, historical faults and maintenance records of equipment, accurately predict the long-term performance of equipment, and make more reasonable selection decisions based on this. At the same time, artificial intelligence technology, especially machine learning and neural networks, can help establish more accurate equipment reliability models, thereby providing more accurate life cycle predictions and cost estimates for equipment selection. By building an evaluation system based on full life cycle costs and equipment reliability scores, a scientific basis can be provided for equipment selection, thereby effectively reducing operational risks and optimizing the management and configuration of power grid assets.

[0004] Prior art, such as the Chinese patent with patent number "CN116664104A", discloses a transformer operating status profiling method based on multi-dimensional capabilities and multi-layer perceptrons, which includes the following steps; Step S1: Collecting equipment information and status parameters of the transformer for building a transformer equipment information label library; Step S2: Building a transformer equipment information label library from N capability dimensions, converting equipment status parameters into massive labeled data, and using it as input for the next algorithm; Step S3: Building a fully connected layer of the multi-layer perceptron algorithm to achieve data compression and feature extraction of equipment status quantities, and realizing transformer status prediction for each capability dimension and the whole through activation functions; Step S4: For transformers in poor condition, locate the faulty part of the equipment according to the scores of each capability dimension and the overall status, and carry out equipment analysis and decision making.

[0005] The problem with the above-mentioned prior art is that when this method constructs a fully connected layer to achieve data compression and feature extraction, the multi-layer perceptron itself is prone to overfitting problems. Especially when faced with a large amount of state data of complex equipment such as transformers, if the amount of training data is relatively insufficient, the model complexity is high, or there is a lack of effective regularization means, it is likely that the model performs well on the training set, but has poor generalization ability when actually applied to other transformer data, resulting in inaccurate state prediction results. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes an equipment selection method and system based on the design of a full life cycle cost portrait system.

[0007] The technical solution of the present invention is as follows:

[0008] The present invention proposes an equipment selection method based on the design of a full life cycle cost portrait system, comprising the following steps:

[0009] Step S1, constructing a life cycle portrait label system for equipment, including a cost label system and a reliability label system for each stage of the life cycle of the equipment;

[0010] Step S2, collecting the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performing outlier detection and missing value interpolation through clustering algorithm and adversarial neural network;

[0011] Step S3, constructing an equipment reliability scoring system based on the equipment reliability label system, and calculating the equipment reliability score;

[0012] Step S4, based on the equipment reliability scoring result, the equipment life cycle portrait system is classified by using the fuzzy C-means clustering method;

[0013] Step S5, based on the image classification results, construct equipment life cycle cost prediction models respectively to obtain the equipment life cycle prediction cost;

[0014] Step S6, calculating the comprehensive score of the equipment through a weighted scoring method according to the equipment reliability score result and the predicted cost of the equipment throughout its life cycle.

[0015] As a preferred implementation, the equipment life cycle portrait label system is constructed, wherein the first-level indicators of the cost label system for each stage of the equipment life cycle include: initial investment cost, operation and maintenance cost, failure and repair cost, and retirement and disposal cost, and a number of second-level and third-level indicators are set under each first-level indicator; the first-level indicators of the reliability label system include: equipment reliability; the second-level indicators include: mean time between failures, equipment life and failure rate; the third-level indicators include: failure interval time, service life and failure probability per unit time.

[0016] As a preferred implementation, outlier detection and missing value interpolation are performed by clustering algorithm and adversarial neural network, wherein: outlier detection is performed by DBSCAN clustering algorithm; missing values ​​in the data are interpolated by adversarial neural network, wherein the adversarial neural network generator loss function is as follows:

[0017] L r (z)=∥x·mG(z)·m∥ 2 ;

[0018] Where: G(z) is the newly generated sample; L r (z) represents the distance between the original data and the newly generated sample; x is the input data of the anti-neural network generator, and m is the reference point of the generator's latent space;

[0019] The discriminator loss function is as follows:

[0020] L d (z) = -D(G(z));

[0021] Where: L d (z) is the discriminator loss function; D() is the discriminator function;

[0022] The filling loss function is obtained through the generator loss function and the discriminator loss function, as follows:

[0023] L imputation (z) = L r (z)+λ·L d (z);

[0024] Where: L imputation (z) is the padding loss function; λ is a hyperparameter;

[0025] The specific method of filling missing values ​​is as follows:

[0026] x imputed =x·m+(1-m)·G(z * );

[0027] Where: x imputed is the sample data after filling; z * The optimal input vector obtained by minimizing the filling function.

[0028] As a preferred implementation, the fuzzy C-means clustering method is used to classify the equipment through the equipment life cycle portrait system, and specifically includes the following steps:

[0029] Confirm that the sample set to be clustered is X. The specific formula is as follows:

[0030] X={x1,x2,x3…x n};

[0031] Where each sample x i There are s characteristic indicators, and the specific formula is as follows:

[0032] x i ={x i1 ,x i2 ,x i3 …x is};

[0033] The sample set to be clustered is divided into C categories. The specific formula is as follows:

[0034] C={c1,c2,c3…c k};

[0035] Where: n is the number of samples to be clustered; s is the number of characteristic indicators; k is the number of categories; i = 1, 2, 3, ..., n;

[0036] The clustering loss function is:

[0037]

[0038] Where: J is the value function of C-means clustering; b is the weighting coefficient; u ij is the membership of the ith sample to cluster j; x i is the i-th sample, with dimension s; c j is the jth cluster, with dimension s; ||x i -c j || represents sample x i With c j The distance to the cluster center;

[0039] The Lagrange multiplier method is used to introduce the equality constraint into the objective function, and the objective function is obtained as follows:

[0040]

[0041] Where: J ‘ is the objective function; i is the Lagrange multiplier of the i-th sample;

[0042] The membership coefficient and cluster center are iteratively calculated until the fuzzy C-means clustering algorithm converges; the convergence condition is as follows:

[0043]

[0044] Where: K is the number of iterations; ε is the error threshold.

[0045] As a preferred implementation manner, the equipment life cycle portrait system is used to classify the equipment through the fuzzy C-means clustering method, and the obtained equipment portrait classification results include: high reliability, low initial investment equipment portrait, high initial investment, high-quality reliability equipment portrait, low reliability, low initial investment equipment portrait, low initial investment, acceptable reliability equipment portrait, high initial investment, high operating cost equipment portrait, high reliability, relatively high operating cost equipment portrait, low reliability, high operating cost equipment portrait, medium investment, balanced reliability and cost equipment portrait, high initial investment, low maintenance requirement equipment portrait, high performance, high maintenance requirement equipment portrait.

[0046] As a preferred implementation, based on the image classification results, equipment life cycle cost prediction models are constructed to obtain equipment life cycle prediction costs, wherein the equipment life cycle cost prediction model is a K-fold-GA-BP neural network model that combines k cross validation with genetic algorithm and back propagation neural network. The specific steps of constructing the K-fold-GA-BP neural network model are as follows:

[0047] K-fold cross validation:

[0048] Divide the data into Q subsets according to the device portrait classification results, and perform Q rounds of verification and training;

[0049] Genetic algorithm GA optimization:

[0050] The genetic algorithm GA is used to determine the initial weights and thresholds of the back propagation BP neural network, which specifically includes the following steps:

[0051] Encode the real number string population consisting of the weights between the input layer, the hidden layer and the output layer, and the thresholds of the hidden layer and the output layer;

[0052] Calculate the fitness of an individual. The specific formula is as follows:

[0053]

[0054] Where: F is the individual fitness; r is the number of output nodes; p w is the expected output value of the w-th node; o w is the expected output value of the w-th node;

[0055] The population is combined according to individual fitness. The specific formula is as follows:

[0056]

[0057] Where: f l is the fitness of the lth population; L is the population size; P l is the probability of population l being selected;

[0058] Randomly select parent individuals for crossover and mutation; use the best individual as the weight and threshold of the back-propagation BP neural network;

[0059] A back propagation BP neural network is constructed according to the optimization result of the genetic algorithm GA. The back propagation BP neural network includes: an input layer, a hidden layer and an output layer:

[0060] Initialize the back propagation BP neural network as follows:

[0061]

[0062] In the formula: A is the neural network neuron input vector, which is the cost label component of each stage of the equipment life cycle; Y is the neural network neuron output vector, which is the predicted cost of each stage of the life cycle; T is the number of input layer nodes; R is the number of output layer nodes;

[0063] The hidden layer expression is as follows:

[0064]

[0065] Where: H yin is the output value of the hidden layer; ∫ is the activation function; λ yin is the hidden layer threshold; ω in is the weight between the hidden layer and the input layer; a t Input value to the neuron; t = 1, 2, 3, ..., T;

[0066] The output layer expression is as follows:

[0067]

[0068] Where: H out is the output value of the output layer;out is the output layer threshold; ω out is the weight between the hidden layer and the output layer; r=1,2,3,…,R.

[0069] As a preferred implementation, the comprehensive score of the equipment is calculated by a weighted scoring method based on the equipment reliability score result and the predicted cost of the equipment throughout its life cycle. The comprehensive score calculation formula is as follows:

[0070] O = α × Z + (1-α) (1-C ′ );

[0071] Where: O is the result of comprehensive score calculation; α is the reliability score weight; C ′ is the standardized equipment life cycle cost.

[0072] On the other hand, the present invention also provides an equipment selection system designed based on a full life cycle cost portrait system, comprising:

[0073] The portrait label system construction module is used to build a portrait label system for the entire life cycle of the equipment, including the cost label system and reliability label system for each stage of the equipment's entire life cycle;

[0074] The data collection and preprocessing module collects the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performs outlier detection and missing value interpolation through clustering algorithm and adversarial neural network;

[0075] The reliability score calculation module builds an equipment reliability score system based on the equipment reliability label system and calculates the equipment reliability score;

[0076] The portrait classification module classifies the equipment portraits based on the equipment reliability score and the fuzzy C-means clustering method for the equipment life cycle portrait system;

[0077] The cost prediction module builds equipment life cycle cost prediction models based on the image classification results to obtain the equipment life cycle prediction cost;

[0078] The comprehensive score calculation module calculates the comprehensive score of the equipment through a weighted scoring method based on the equipment reliability score results and the predicted cost of the equipment throughout its life cycle.

[0079] On the other hand, the present invention further provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for selecting equipment based on the design of a full life cycle cost portrait system as described in any embodiment of the present invention is implemented.

[0080] On the other hand, the present invention also provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement an equipment selection method based on the design of a full life cycle cost portrait system as described in any embodiment of the present invention.

[0081] The present invention has the following beneficial effects:

[0082] 1. Comprehensively consider the cost and reliability of the entire life cycle of the equipment, avoid focusing only on the subsequent high costs caused by the initial investment cost, provide a more scientific and comprehensive basis for equipment selection, and help improve the economic benefits and long-term reliability of the power grid system.

[0083] 2. The DBSCAN clustering algorithm is used to effectively detect outliers in the data set, ensuring the accuracy and reliability of the data and providing a high-quality data foundation for subsequent analysis. The adversarial neural network is used to interpolate missing values, and the generator loss function and the discriminator loss function are combined to obtain the filling loss function, so that the interpolated data is closer to the true value and improves data integrity.

[0084] 3. The fuzzy C-means clustering method is used to classify the equipment portraits and obtain various types of portraits, which helps to gain a deeper understanding of the equipment characteristics and provide more targeted guidance for equipment selection in different demand scenarios.

[0085] 4. The K-fold-GA-BP neural network model constructed based on the portrait classification results can effectively improve the accuracy of cost prediction through K-fold cross-validation, genetic algorithm optimization and back-propagation neural network training, providing strong support for the cost assessment of the entire life cycle of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0087] Figure 1 This is a schematic diagram of the method flow of Example 1; DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0090] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0091] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0092] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0093] Embodiment 1:

[0094] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following will be combined with the specific embodiments of the present application and refer to the attached Figure 1 , clearly and completely describe the technical solution of the present invention.

[0095] To solve the problems of the prior art, the present invention provides an equipment selection method based on a full life cycle cost portrait system design, comprising the following steps:

[0096] Step S1, constructing a life cycle portrait label system for equipment, including a cost label system and a reliability label system for each stage of the life cycle of the equipment;

[0097] The equipment life cycle portrait label system is constructed, wherein the first-level indicators of the cost label system at each stage of the equipment life cycle include: initial investment cost, operation and maintenance cost, failure and repair cost, and retirement and disposal cost, and several second-level and third-level indicators are set under each first-level indicator; the first-level indicator of the reliability label system includes: equipment reliability; the second-level indicators include: mean time between failures, equipment life and failure rate; the third-level indicators include: failure interval time, service life and failure probability per unit time. The specific portrait label system is shown in Table 1;

[0098] Table 1 Equipment life cycle portrait label system

[0099]

[0100]

[0101] Step S2, collecting the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performing outlier detection and missing value interpolation through clustering algorithm and adversarial neural network;

[0102] Step S201, performing outlier detection using the DBSCAN clustering algorithm;

[0103] Comprehensively collect data related to the entire life cycle of the equipment. This data includes the initial investment cost of the equipment, operation and maintenance costs, failure and repair records, and the retirement and disposal costs of the equipment. In addition, it is necessary to collect equipment usage environment data and equipment reliability data. These data provide a scientific basis for subsequent cost prediction and equipment selection optimization.

[0104] Set the neighborhood radius and the minimum number of points; take each data point from the data set in turn, calculate the number of all data points contained in its neighborhood radius, if the number of all data points contained is ≥ the minimum number of points, then the data point is a core point, all points in the neighborhood of the data point (if these points have not been divided into other clusters) are included in this cluster, and repeat this step for these newly included points, and continue to expand the cluster until there are no more points that can be expanded in the cluster (that is, the points in the neighborhood are either already in other clusters, or the number of points in its own neighborhood is < the minimum number of points; if the number of all data points contained in the neighborhood radius of the data point is ≤ the minimum number of points, it is temporarily marked as a non-core point; after the cluster expansion is completed, those points that do not belong to any core point neighborhood but are in the neighborhood of a cluster core point are marked as boundary points, and they belong to the corresponding cluster. And those data points that are neither core points nor included in any cluster as boundary points are judged as noise points (i.e. outliers)

[0105] Step S202, interpolating missing values ​​in the data by using an adversarial neural network;

[0106] Among them, the loss function of the adversarial neural network generator is as follows:

[0107] L r (z)=∥x·mG(z)·m∥ 2 ;

[0108] Where: G(z) is the newly generated sample; L r(z) represents the distance between the original data and the newly generated sample; x is the input data of the anti-neural network generator, and m is the reference point of the generator's latent space;

[0109] The discriminator loss function is as follows:

[0110] L d (z) = -D(G(z));

[0111] Where: L d (z) is the discriminator loss function; D() is the discriminator function;

[0112] The filling loss function is obtained through the generator loss function and the discriminator loss function, as follows:

[0113] L imputation (z) = L r (z)+λ·L d (z);

[0114] Where: L imoutation (z) is the padding loss function; λ is a hyperparameter;

[0115] The specific method of filling missing values ​​is as follows:

[0116] x imputed =x·m+(1-m)·G(z * );

[0117] Where: x imputed is the sample data after filling; z * The optimal input vector obtained by minimizing the filling function.

[0118] Step S3, constructing an equipment reliability scoring system based on the equipment reliability label system, and calculating the equipment reliability score;

[0119] The construction of the reliability scoring system includes the following steps:

[0120] Reliability indicators: First, through the data collection and profiling results in the previous steps, key indicators related to equipment reliability are extracted, such as mean time between failures, failure rate, maintenance frequency, fault repair time, and equipment life.

[0121] Data standardization: In order to ensure the comparability of various indicators during calculation, the collected reliability data needs to be standardized. Through the normalization method, indicators with different units and dimensions are converted into a unified scale.

[0122] Weighted scoring: Each reliability indicator is given a certain weight, and the weight can be determined by the analytic hierarchy process. The reliability scores of different devices are calculated based on these indicators and weights to obtain a comprehensive reliability score for each device.

[0123] In order to reflect the subjective emphasis on the three links before, during and after the event and to determine the scores of the indicators before and during the event, the present invention adopts the hierarchical analysis method to quantify the subjective emphasis.

[0124] Decompose complex decision-making problems into different levels such as goals, criteria (or indicators), sub-criteria (if necessary) and alternative plans, and form a multi-level structural model, including the goal level, criterion level, sub-criteria level (if applicable) and plan level. In each level, the elements are compared in pairs, and the 1-9 scale method is used to evaluate their relative importance. According to the results of the pairwise comparison, a judgment matrix is ​​constructed, and the relative weight of each element is calculated by solving the eigenvector of the judgment matrix. Finally, the total weight of the plan level relative to the goal level is calculated, and the optimal plan is determined by sorting, and the decision is made based on the total weight to select the optimal plan.

[0125] The steps for determining the score using the hierarchical analysis method are as follows:

[0126] Determine the indicators and ultimate goals required for the ex ante evaluation system;

[0127] Construct a hierarchical structure: the first layer is the target layer, the second layer is the first-level indicator layer, and the third layer is the second-level indicator layer;

[0128] Constructing a judgment matrix: Based on the experts' evaluation of the relative importance of each indicator using a 1-9 scale, a judgment matrix at the target and indicator levels is constructed;

[0129] Calculate the corresponding weights of each layer element: obtain the weight of each element by solving the eigenvector of the judgment matrix. ;

[0130] By using the weighted average method, each standardized reliability index is multiplied by its corresponding weight, and then the sum is calculated to obtain the comprehensive reliability score of the equipment.

[0131] Step S4, based on the equipment reliability scoring result, the equipment life cycle portrait system is classified by using the fuzzy C-means clustering method;

[0132] The fuzzy C-means clustering method is used to classify the equipment through the equipment life cycle portrait system, which specifically includes the following steps:

[0133] Confirm that the sample set to be clustered is X. The specific formula is as follows:

[0134] X={x1,x2,x3…x n};

[0135] Where each sample x i There are s characteristic indicators, and the specific formula is as follows:

[0136] x i ={x i1 ,x i2 ,x i3 …x is};

[0137] The sample set to be clustered is divided into C categories. The specific formula is as follows:

[0138] C={c1,c2,c3…c k};

[0139] Where: n is the number of samples to be clustered; s is the number of characteristic indicators; k is the number of categories; i = 1, 2, 3, ..., n;

[0140] The clustering loss function is:

[0141]

[0142] Where: J is the value function of C-means clustering; b is the weighting coefficient; u ij is the membership of the ith sample to cluster j; x i is the i-th sample, with dimension s; c j is the jth cluster, with dimension s; ||x i -c j || represents sample x i With c j The distance to the cluster center;

[0143] The Lagrange multiplier method is used to introduce the equality constraint into the objective function, and the objective function is obtained as follows:

[0144]

[0145] The membership calculation formula is:

[0146]

[0147] The cluster center calculation formula is:

[0148]

[0149] Where: J ‘ is the objective function; i is the Lagrange multiplier of the i-th sample;

[0150] The membership coefficient and cluster center are iteratively calculated until the fuzzy C-means clustering algorithm converges; the convergence condition is as follows:

[0151]

[0152] Where: K is the number of iterations; ε is the error threshold.

[0153] The equipment portrait classification results are as follows: high reliability and low initial investment equipment portrait, high initial investment and high-quality reliability equipment portrait, low reliability and low initial investment equipment portrait, low initial investment and acceptable reliability equipment portrait, high initial investment and high operating cost equipment portrait, high reliability and relatively high operating cost equipment portrait, low reliability and high operating cost equipment portrait, medium investment and balanced reliability and cost equipment portrait, high initial investment and low maintenance requirement equipment portrait, and high performance and high maintenance requirement equipment portrait.

[0154] Step S5, based on the image classification results, construct equipment life cycle cost prediction models respectively to obtain the equipment life cycle prediction cost;

[0155] The equipment life cycle cost prediction model is a K-fold-GA-BP neural network model that combines k cross validation with genetic algorithm and back propagation neural network. The specific steps of constructing the K-fold-GA-BP neural network model are as follows:

[0156] K-fold cross validation:

[0157] Divide the data into Q subsets according to the device portrait classification results, and perform Q rounds of verification and training;

[0158] Genetic algorithm GA optimization:

[0159] The genetic algorithm GA is used to determine the initial weights and thresholds of the back propagation BP neural network, which specifically includes the following steps:

[0160] Encode the real number string population consisting of the weights between the input layer, the hidden layer and the output layer, and the thresholds of the hidden layer and the output layer;

[0161] Calculate the fitness of an individual. The specific formula is as follows:

[0162]

[0163] Where: F is the individual fitness; r is the number of output nodes; p w is the expected output value of the w-th node; o w is the expected output value of the w-th node;

[0164] The population is combined according to individual fitness. The specific formula is as follows:

[0165]

[0166] Where: f l is the fitness of the lth population; L is the population size; P l is the probability of population l being selected;

[0167] Randomly select parent individuals for crossover and mutation; use the best individual as the weight and threshold of the back-propagation BP neural network;

[0168] A back propagation BP neural network is constructed according to the optimization result of the genetic algorithm GA. The back propagation BP neural network includes: an input layer, a hidden layer and an output layer:

[0169] Initialize the back propagation BP neural network as follows:

[0170]

[0171] In the formula: A is the neural network neuron input vector, which is the cost label component of each stage of the equipment life cycle; Y is the neural network neuron output vector, which is the predicted cost of each stage of the life cycle; T is the number of input layer nodes; R is the number of output layer nodes;

[0172] The hidden layer expression is as follows:

[0173]

[0174] Where: H yin is the output value of the hidden layer; ∫ is the activation function; λ yin is the hidden layer threshold; ω in is the weight between the hidden layer and the input layer; a t Input value to the neuron; t = 1, 2, 3, ..., T;

[0175] The output layer expression is as follows:

[0176]

[0177] Where: H out is the output value of the output layer; out is the output layer threshold; ω out is the weight between the hidden layer and the output layer; r=1,2,3,…,R.

[0178] Let e0 be the maximum allowable error value. If max(e)>e0, adjust the BP neural network parameters and re-train until all training data meet the specified error standard.

[0179] The error calculation formula is as follows:

[0180] e=|Y out -H out |;

[0181] Where: e is the error value; Y out is the expected output value;

[0182] According to the error value, the weights between the hidden layer and the input layer, the weights between the hidden layer and the output layer, the hidden layer threshold and the output layer threshold are updated; the specific formula is as follows:

[0183]

[0184] ω ′ out =ω out +ηH yin e;

[0185]

[0186] λ ′ out =λ out +e;

[0187] Where: i ′ n ,ω ′ out , ′ yin and λ ′ out are the weights between the hidden layer and the input layer after error update, the weights between the hidden layer and the output layer, the hidden layer threshold, and the output layer threshold; η is the learning rate.

[0188] The mean absolute percentage error and root mean square error method can be used to evaluate the impact of K-fold cross validation on model accuracy and the effectiveness of BP neural network prediction.

[0189] Step S6, calculating the comprehensive score of the equipment through a weighted scoring method according to the equipment reliability score result and the predicted cost of the equipment throughout its life cycle.

[0190] A simplified method based on weighted scoring is adopted to combine the life cycle cost and reliability score of the equipment to provide effective decision support for equipment selection. First, the life cycle cost of each equipment is calculated, covering the initial investment, operation and maintenance, fault repair and decommissioning costs; then the reliability score is calculated based on the reliability index of the equipment; finally, the life cycle cost is combined with the reliability score through the weighted scoring method to obtain a comprehensive score for each equipment, which guides the equipment selection and optimization decision, ensuring the long-term stability and reliability of the equipment while meeting the cost-effectiveness.

[0191] When selecting a model, the life cycle cost and reliability score of the equipment are combined to form a comprehensive scoring model. The specific formula is as follows:

[0192] O = α × Z + (1-α) (1-C ′ );

[0193] Where: O is the result of comprehensive score calculation; α is the reliability score weight; C ′ is the standardized equipment life cycle cost.

[0194] The cost and reliability scores of each device can be combined to obtain a comprehensive score for the devices and sort them. The device with the higher score is the preferred device in the selection process.

[0195] Embodiment 2:

[0196] This embodiment provides an equipment selection system designed based on a full life cycle cost profile system, including:

[0197] The portrait label system construction module is used to build a portrait label system for the entire life cycle of the equipment, including the cost label system and reliability label system for each stage of the equipment's entire life cycle;

[0198] The data collection and preprocessing module collects the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performs outlier detection and missing value interpolation through clustering algorithm and adversarial neural network;

[0199] The reliability score calculation module builds an equipment reliability score system based on the equipment reliability label system and calculates the equipment reliability score;

[0200] The portrait classification module classifies the equipment portraits based on the equipment reliability score and the fuzzy C-means clustering method for the equipment life cycle portrait system;

[0201] The cost prediction module builds equipment life cycle cost prediction models based on the image classification results to obtain the equipment life cycle prediction cost;

[0202] The comprehensive score calculation module calculates the comprehensive score of the equipment through a weighted scoring method based on the equipment reliability score results and the predicted cost of the equipment throughout its life cycle.

[0203] Embodiment three:

[0204] This embodiment provides an electronic device having a computer program stored thereon. When the computer program is executed by a processor, the method for selecting equipment based on a full life cycle cost portrait system design as described in any embodiment of the present invention is implemented.

[0205] Embodiment 4:

[0206] This embodiment provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement an equipment selection method based on a full life cycle cost portrait system design as described in any embodiment of the present invention.

[0207] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0208] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0209] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0210] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0211] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for equipment selection based on the design of a full life cycle cost portrait system, characterized in that: The following steps are involved: Step S1, constructing a life cycle portrait label system for equipment, including a cost label system and a reliability label system for each stage of the life cycle of the equipment; Step S2, collecting the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performing outlier detection and missing value interpolation through clustering algorithm and adversarial neural network; Step S3, constructing an equipment reliability scoring system based on the equipment reliability label system, and calculating the equipment reliability score; Step S4, based on the equipment reliability scoring result, the equipment life cycle portrait system is classified by using the fuzzy C-means clustering method; Step S5, based on the image classification results, construct equipment life cycle cost prediction models respectively to obtain the equipment life cycle prediction cost; Step S6, calculating the comprehensive score of the equipment through a weighted scoring method according to the equipment reliability score result and the predicted cost of the equipment throughout its life cycle.

2. The equipment selection method based on the full life cycle cost portrait system design according to claim 1 is characterized by: The equipment life cycle portrait label system is constructed, wherein the first-level indicators of the cost label system for each stage of the equipment life cycle include: initial investment cost, operation and maintenance cost, failure and repair cost, and retirement and disposal cost, and a number of second-level and third-level indicators are set under each first-level indicator; the first-level indicators of the reliability label system include: equipment reliability; the second-level indicators include: average failure interval time, equipment life and failure rate; the third-level indicators include: failure interval time, service life and failure probability per unit time.

3. The equipment selection method based on the full life cycle cost portrait system design according to claim 1 is characterized by: The method uses a clustering algorithm and an adversarial neural network to detect outliers and interpolate missing values, wherein: outlier detection is performed using a DBSCAN clustering algorithm; missing values ​​in the data are interpolated using an adversarial neural network, wherein the adversarial neural network generator loss function is as follows: L r (z)=||x·m-G(z)·m|| 2 ; Where: G(z) is the newly generated sample; L r (z) represents the distance between the original data and the newly generated sample; x is the input data of the anti-neural network generator, and m is the reference point of the generator's latent space; The discriminator loss function is as follows: L d (z)=-D(G(z)); Where: L d (z) is the discriminator loss function; D() is the discriminator function; The filling loss function is obtained through the generator loss function and the discriminator loss function, as follows: L imputation (from)=L r (z)+λ L d (With); Where: L imputation (z) is the padding loss function; λ is a hyperparameter; The specific method of filling missing values ​​is as follows: x imputed =x·m+(1-m)·G(z * ); Where: x imputed is the sample data after filling; z * The optimal input vector obtained by minimizing the filling function.

4. The equipment selection method based on the full life cycle cost portrait system design according to claim 1 is characterized by: The device life cycle portrait system is used to classify the device using the fuzzy C-means clustering method, and the following steps are specifically included: Confirm that the sample set to be clustered is X. The specific formula is as follows: X={x1,x2,x3…x n }; Where each sample x i There are s characteristic indicators, and the specific formula is as follows: x i ={x i1 ,x i2 ,x i3 …x is }; The sample set to be clustered is divided into C categories. The specific formula is as follows: C={c1,c2,c3…c k }; Where: n is the number of samples to be clustered; s is the number of characteristic indicators; k is the number of categories; i = 1, 2, 3, ..., n; The clustering loss function is: Where: J is the value function of C-means clustering; b is the weighting coefficient; u ij is the membership of the ith sample to cluster j; x i is the i-th sample, with dimension s; c j is the jth cluster, with dimension s; ||x i -c j || represents sample x i With c j The distance to the cluster center; The Lagrange multiplier method is used to introduce the equality constraint into the objective function, and the objective function is obtained as follows: Where: J' is the objective function; λ i is the Lagrange multiplier of the i-th sample; The membership coefficient and cluster center are iteratively calculated until the fuzzy C-means clustering algorithm converges; the convergence condition is as follows: Where: K is the number of iterations; ε is the error threshold.

5. The equipment selection method based on the full life cycle cost portrait system design according to claim 4 is characterized by: The equipment portrait classification system of the equipment life cycle is carried out through fuzzy C-means clustering method, and the obtained equipment portrait classification results include: high reliability, low initial investment equipment portrait, high initial investment, high-quality reliability equipment portrait, low reliability, low initial investment equipment portrait, low initial investment, acceptable reliability equipment portrait, high initial investment, high operating cost equipment portrait, high reliability, relatively high operating cost equipment portrait, low reliability, high operating cost equipment portrait, medium investment, balanced reliability and cost equipment portrait, high initial investment, low maintenance requirement equipment portrait, high performance, high maintenance requirement equipment portrait.

6. The equipment selection method based on the full life cycle cost portrait system design according to claim 1 is characterized by: Based on the image classification results, equipment life cycle cost prediction models are constructed to obtain equipment life cycle prediction costs. The equipment life cycle cost prediction model is a K-fold-GA-BP neural network model that combines k cross validation with genetic algorithm and back propagation neural network. The specific steps of constructing the K-fold-GA-BP neural network model are as follows: K-fold cross validation: Divide the data into Q subsets according to the device portrait classification results, and perform Q rounds of verification and training; Genetic algorithm GA optimization: The genetic algorithm GA is used to determine the initial weights and thresholds of the back propagation BP neural network, which specifically includes the following steps: Encode the real number string population consisting of the weights between the input layer, the hidden layer and the output layer, and the thresholds of the hidden layer and the output layer; Calculate the fitness of an individual. The specific formula is as follows: Where: F is the individual fitness; r is the number of output nodes; p w is the expected output value of the w-th node; o w is the expected output value of the w-th node; The population is combined according to individual fitness. The specific formula is as follows: Where: f l is the fitness of the lth population; L is the population size; P l is the probability of population l being selected; Randomly select parent individuals for crossover and mutation; use the best individual as the weight and threshold of the back-propagation BP neural network; A back propagation BP neural network is constructed according to the optimization result of the genetic algorithm GA. The back propagation BP neural network includes: an input layer, a hidden layer and an output layer: Initialize the back propagation BP neural network as follows: In the formula: A is the neural network neuron input vector, which is the cost label component of each stage of the equipment life cycle; Y is the neural network neuron output vector, which is the predicted cost of each stage of the life cycle; T is the number of input layer nodes; R is the number of output layer nodes; The hidden layer expression is as follows: Where: H yin is the output value of the hidden layer; ∫ is the activation function; λ yin is the hidden layer threshold; ω in is the weight between the hidden layer and the input layer; a t Input value to the neuron; t = 1, 2, 3, ..., T; The output layer expression is as follows: Where: H out is the output value of the output layer; out is the output layer threshold; ω out is the weight between the hidden layer and the output layer; r=1,2,3,…,R.

7. The equipment selection method based on the full life cycle cost portrait system design according to claim 1 is characterized by: According to the equipment reliability score result and the predicted cost of the equipment throughout its life cycle, the comprehensive score of the equipment is calculated by weighted scoring method. The comprehensive score calculation formula is as follows: O=α×Z+(1-α)(1-C ′ ); Where: O is the result of comprehensive score calculation; α is the reliability score weight; C ′ is the standardized equipment life cycle cost.

8. An equipment selection system based on the design of the full life cycle cost portrait system, characterized in that: include: The portrait label system construction module is used to build a portrait label system for the entire life cycle of the equipment, including the cost label system and reliability label system for each stage of the equipment's entire life cycle; The data collection and preprocessing module collects the corresponding equipment life cycle cost data and reliability data according to the equipment life cycle portrait label system, and performs outlier detection and missing value interpolation through clustering algorithm and adversarial neural network; The reliability score calculation module builds an equipment reliability score system based on the equipment reliability label system and calculates the equipment reliability score; The portrait classification module classifies the equipment portraits based on the equipment reliability score and the fuzzy C-means clustering method for the equipment life cycle portrait system; The cost prediction module builds equipment life cycle cost prediction models based on the image classification results to obtain the equipment life cycle prediction cost; The comprehensive score calculation module calculates the comprehensive score of the equipment through a weighted scoring method based on the equipment reliability score results and the predicted cost of the equipment throughout its life cycle.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements an equipment selection method based on the design of a full life cycle cost portrait system as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements an equipment selection method based on a full life cycle cost portrait system design as described in any one of claims 1 to 7.

Images