Method and device for determining maintenance cost data of transformer substation

By combining the target deep confidence network, the target backpropagation neural network and the target gray prediction model, quantile regression average processing is used to solve the problem of insufficient reliability in determining the cost data of the substation, and the reliability of the prediction results of the calculation model is improved. .

CN119067692APending Publication Date: 2024-12-03ZHANJIANG POWER SUPPLY BUREAU OF GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411044272.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the prior art, there is a problem of insufficient reliability in the determination of substation maintenance cost data, resulting in inaccurate calculation results.

Method used

The target deep confidence network, target backpropagation neural network and target gray prediction model are used to calculate the cost impact data, and multiple calculation results are fused through quantile regression average processing to obtain the target cost data.

Benefits of technology

The reliability of the prediction results of substation maintenance costs in the actual project is improved, and the problem of insufficient reliability in the determination of cost data is solved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067692A_ABST
    Figure CN119067692A_ABST
Patent Text Reader

Abstract

The invention provides a method and a device for determining maintenance cost data of a transformer substation, and the method comprises the steps: obtaining cost influence data, the cost influence data being influence data which has influence on cost data and has the influence degree at the top m, and the cost data being the maintenance cost of primary equipment of the transformer substation, the cost influence data comprises at least one of engineering parameters of the transformer substation, equipment parameters of the transformer substation and process parameters of the transformer substation; calculating the cost influence data by using a predetermined model to obtain a plurality of calculation results, the predetermined model including a target deep belief network, a target back propagation neural network and a target grey prediction model; and performing fusion processing on the plurality of calculation results to obtain target cost data, the fusion processing at least comprising quantile regression average processing. The method solves the problem of insufficient reliability in the process of determining the cost data in the prior art.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of substation maintenance, and in particular, to a method for determining substation maintenance cost data, a device for determining substation maintenance cost data, a computer-readable storage medium, and a computer program product. Background Art

[0002] In the electric power industry, substation maintenance is an important part of maintaining the normal operation of substation equipment and ensuring the safety and stability of users' electricity use. As electricity demand increases, the operation and maintenance costs of substation projects are also increasing. Therefore, the existing cost estimation methods have defects and shortcomings in terms of accuracy and reliability.

[0003] Therefore, there is an urgent need for a method to solve the problem of insufficient reliability in the process of determining cost data in the prior art. Summary of the invention

[0004] The main purpose of the present application is to provide a method for determining substation maintenance cost data, a device for determining substation maintenance cost data, a computer-readable storage medium and a computer program product, so as to at least solve the problem of insufficient reliability in the process of determining cost data in the prior art.

[0005] According to one aspect of the present application, a method for determining substation maintenance cost data is provided, and cost impact data is obtained, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the top m positions, the cost data is the maintenance cost of the primary equipment of the substation, and the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation and the process parameters of the substation, and m is a positive integer greater than 0; a predetermined model is constructed, and the cost impact data is calculated using the predetermined model to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network and a target grey prediction model, the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one; a plurality of the calculation results are fused to obtain target cost data, wherein the fusion processing at least includes quantile regression averaging processing.

[0006] Optionally, constructing the target deep belief network includes: obtaining an initial deep belief network, wherein the initial deep belief network includes multiple restricted Boltzmann machines and back propagation neural networks, the multiple restricted Boltzmann machines are used to extract features of the input data of the initial deep belief network layer by layer, and the back propagation neural network is used for regression prediction; using an unsupervised learning algorithm to train the weights of the initial deep belief network and the bias of the initial deep belief network layer by layer to pre-train the initial deep belief network; using forward propagation, back propagation and supervised learning algorithms to process the initial deep belief network to train the initial deep belief network; obtaining the mean absolute percentage error and root mean square error of the initial deep belief network, and when the mean absolute percentage error is less than a first error threshold and the root mean square error is less than a second error threshold, determining that the initial deep belief network is the target deep belief network.

[0007] Optionally, constructing the target back propagation neural network includes: obtaining an initial back propagation neural network, and using a genetic algorithm to process initial parameters of the back propagation neural network to obtain target parameters, wherein the initial parameters include an initial threshold and an initial weight; bringing the target parameters into the initial back propagation neural network to obtain the target back propagation neural network.

[0008] Optionally, the target back propagation neural network is used to calculate the cost impact data to obtain multiple calculation results, including: obtaining multiple historical impact data and multiple historical cost data, wherein the historical impact data is the cost impact data acquired before the current moment, and the historical cost data is the cost data acquired before the current moment; inputting the cost impact data into the target back propagation neural network to obtain calculation results corresponding to the cost impact data, wherein the target back propagation neural network is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes the historical impact data and the historical cost data corresponding to the historical impact data.

[0009] Optionally, constructing the target grey prediction model includes: obtaining a plurality of historical impact data and a plurality of historical cost data, wherein the historical impact data is the cost impact data obtained before the current moment, and the historical cost data is the cost data obtained before the current moment; according to the formula Construct the target grey prediction model, μ is the endogenous control grey number, a is the development grey number, k is a positive integer greater than or equal to 0, X (0) (1) is the first data set X (0 ), The second data set X(1) The first data group includes a plurality of the historical impact data arranged in time, and the second data group is generated by accumulating the first data groups.

[0010] Optionally, a plurality of the calculation results are fused to obtain target cost data, including: according to the formula Calculate preliminary cost data in, is the predicted value of the target deep belief network at the nth sample point, Back propagate the predicted value of the target neural network at the nth sample point, is the predicted value of the target grey prediction model at the nth sample point, and the sample points correspond to the cost influencing data one by one; according to the formula and Calculate the loss function L, where q is the quantile, I is the indicator function, and y n is the actual cost data corresponding to the nth sample point, β q is the parameter vector of the quantile q; determine the β corresponding to the minimum function value of the loss function q is the target parameter vector β q ′; According to the formula and Calculate the upper limit value U n , where 1-α is the confidence level; according to the formula and Calculate the lower limit value L n ; Calculate the target cost data based on the upper limit value and the lower limit value.

[0011] Optionally, calculating the target cost data according to the upper limit value and the lower limit value includes: determining a preliminary interval according to the upper limit value and the lower limit value; and calculating the target cost data according to the formula Calculate the credibility of the preparatory interval PICP, wherein, when the preparatory cost data corresponding to the nth sample point belongs to the preparatory interval, c n is 1, and the preliminary cost data corresponding to the nth sample point does not belong to the preliminary interval, c n is 0, N is the total number of sample points; according to the formula: Calculating an average bandwidth PINAW of the preparation interval; and determining the preparation interval as the target cost data when the credibility of the preparation interval is greater than a credibility threshold and the average bandwidth of the preparation interval is greater than a width threshold.

[0012] According to another aspect of the present application, a device for determining substation maintenance cost data is provided, comprising: an acquisition unit, used to acquire cost impact data, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the top m positions, the cost data is the maintenance cost of the primary equipment of the substation, and the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation and the process parameters of the substation, and m is a positive integer greater than 0; a calculation unit, used to construct a predetermined model, and use the predetermined model to calculate the cost impact data to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network and a target grey prediction model, the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one; a processing unit, used to fuse the multiple calculation results to obtain target cost data, wherein the fusion processing at least includes quantile regression averaging processing.

[0013] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0014] According to another aspect of the present application, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any one of the methods described are implemented.

[0015] Applying the technical solution of the present application, a method for determining the maintenance cost data of a substation is provided. First, the cost impact data is obtained. The cost impact data is the impact data that has an impact on the cost data and the impact degree is in the top m positions. The cost data is the maintenance cost of the primary equipment of the substation. The cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation, and the process parameters of the substation; and a predetermined model is used to calculate the cost impact data to obtain multiple calculation results. The predetermined model includes a target deep confidence network, a target back propagation neural network, and a target gray prediction model; the multiple calculation results are fused to obtain the target cost data, and the fusion processing at least includes quantile regression averaging. By combining the prediction results of three different base models, namely the target deep confidence network, the target back propagation neural network, and the target gray prediction model, and fusing the three calculation results by quantile regression averaging, the maintenance cost of the primary equipment of the substation is predicted, so as to improve the reliability of the prediction results of the measurement model in engineering practice, and solve the problem of insufficient reliability in the process of determining cost data in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a mobile terminal for executing a method for determining substation maintenance cost data provided in an embodiment of the present application is shown;

[0018] Figure 2 A schematic flow chart of a method for determining substation maintenance cost data provided according to an embodiment of the present application is shown;

[0019] Figure 3 A schematic diagram of the composition of cost data provided according to an embodiment of the present application is shown;

[0020] Figure 4 A schematic diagram of the structure of a predetermined model provided in an embodiment of the present application is shown;

[0021] Figure 5 A schematic diagram of the structure of a target deep confidence network provided according to an embodiment of the present application is shown;

[0022] Figure 6 A schematic diagram of a process flow of a contrastive divergence learning process provided according to an embodiment of the present application is shown;

[0023] Figure 7 A schematic diagram of the structure of a target back propagation neural network provided in an embodiment of the present application is shown;

[0024] Figure 8 A specific flow chart of a method for determining substation maintenance cost data provided in accordance with an embodiment of the present application is shown;

[0025] Fig. 9 A structural block diagram of a device for determining substation maintenance cost data provided according to an embodiment of the present application is shown.

[0026] The above drawings include the following reference numerals:

[0027] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] As introduced in the background technology, the reliability of the process of determining cost data in the prior art is insufficient. To solve the above problem, the embodiments of the present application provide a method for determining substation maintenance cost data, a device for determining substation maintenance cost data, a computer-readable storage medium and a computer program product.

[0032] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0033] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 1 is a hardware structure block diagram of a mobile terminal for determining a method for substation maintenance cost data according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1Different configurations are shown.

[0034] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the maintenance cost data of the substation in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific examples of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0035] In this embodiment, a method for determining substation maintenance cost data running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] Figure 2 1 is a flow chart of a method for determining substation maintenance cost data according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0037] Step S201, obtaining cost impact data, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the first m positions, the cost data is the maintenance cost of the primary equipment of the substation, the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation and the process parameters of the substation, and m is a positive integer greater than 0;

[0038] Specifically, the substation primary equipment maintenance project consists of multiple processes, and the total cost of the project is the sum of all process costs. The specific cost data is composed of Figure 3 As shown. Due to the complex composition and wide range of process costs, a lot of business logic is required when using traditional methods to make cost budgets, especially the quota type, which requires detailed quota files for data such as the unit price of manpower, machine and materials. The workload is huge and there is a greater risk of error. Therefore, it is necessary to analyze the sample data to identify the key influencing factors of the project cost as the basic data support for establishing the cost estimation model. By sorting out and analyzing the historical settlement audit report data of the substation primary equipment maintenance project, it is concluded that the same type of projects usually have a large number of the same processes. In the substation primary equipment maintenance project, the project parameters and equipment parameters determine the fixed process and its project volume. Therefore, the project parameters, equipment parameters, variable process parameters, etc. can be the key influencing factors of the project cost.

[0039] Step S202, constructing a predetermined model, and using the predetermined model to calculate the cost impact data to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network, and a target grey prediction model, and the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one;

[0040] Specifically, the structural diagram of the above predetermined model is as follows: Figure 4 As shown, by building a machine learning-based engineering cost estimation model framework, namely the above-mentioned predetermined model, an input and output sample set is established according to historical engineering data to obtain the fitting relationship between engineering parameters, equipment parameters, process parameters and engineering costs.

[0041] Step S203, fusing the plurality of the above calculation results to obtain target cost data, wherein the above fusion processing at least includes quantile regression averaging processing.

[0042] Specifically, quantile regression is a statistical technique for estimating conditional quantiles. Unlike common mean regression, quantile regression can provide more comprehensive information about the conditional distribution of the response variable, not just the mean. In traditional linear regression, the focus is on how the expected value (or mean) of the dependent variable changes as the independent variable changes. However, in some cases, other quantiles (such as the median or quartiles) may also be of interest. Quantiles can estimate these conditional quantiles to obtain a deeper understanding of the data distribution. Quantile regression mean treatment refers to a special case of using quantile regression to estimate the conditional mean. In this case, a quantile of 0.5, that is, the median, can be selected for regression analysis. Since the median is insensitive to outliers, this method can provide a robust mean estimate.

[0043] Through this embodiment, a method for determining the maintenance cost data of a substation is provided. First, the cost impact data is obtained. The cost impact data is the impact data that has an impact on the cost data and the impact degree is in the top m positions. The cost data is the maintenance cost of the primary equipment of the substation. The cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation, and the process parameters of the substation; and a predetermined model is used to calculate the cost impact data to obtain multiple calculation results. The predetermined model includes a target deep confidence network, a target back propagation neural network, and a target gray prediction model; the multiple calculation results are fused to obtain the target cost data, and the fusion processing at least includes quantile regression averaging. By combining the prediction results of three different base models, namely the target deep confidence network, the target back propagation neural network, and the target gray prediction model, and fusing the three calculation results by quantile regression averaging, the maintenance cost of the primary equipment of the substation is predicted, so as to improve the reliability of the prediction results of the measurement model in engineering practice, and solve the problem of insufficient reliability in the process of determining cost data in the prior art.

[0044] In the specific implementation process, the above step S202 can be implemented by the following steps: step S2021, obtaining an initial deep belief network, wherein the above initial deep belief network includes multiple restricted Boltzmann machines and back propagation neural networks, the multiple restricted Boltzmann machines are used to extract the features of the input data of the above initial deep belief network layer by layer, and the above back propagation neural network is used for regression prediction; step S2022, using an unsupervised learning algorithm to train the weights of the above initial deep belief network and the bias of the above initial deep belief network layer by layer to pre-train the above initial deep belief network; step S2023, using forward propagation, back propagation and supervised learning algorithms to process the above initial deep belief network to train the above initial deep belief network; step S2024, obtaining the average absolute percentage error and root mean square error of the above initial deep belief network, and determining that the above initial deep belief network is the above target deep belief network when the above average absolute percentage error is less than the first error threshold and the above root mean square error is less than the second error threshold. This method can further quickly establish the above target deep belief network.

[0045] Specifically, the deep belief network (DBN) is an efficient unsupervised learning algorithm proposed by Hinton in 2006. It can be regarded as a combination of a series of restricted Boltzmann machines (RBM). Through the integration of deep learning and feature learning, DBN has good data fitting ability and fast analysis ability. The initial parameters of DBN can be obtained layer by layer through unsupervised pre-training, thereby effectively solving the related problems caused by the random initialization parameters of traditional neural networks. On the other hand, pre-training can effectively avoid overfitting and underfitting problems. The structure of the DBN prediction model is as follows: Figure 5 As shown in the figure, the DBN model consists of multiple RBMs and a single-layer back-propagation neural network (BP neural network for short). The BP neural network has 1 input layer, 1 hidden layer and 1 output layer. The input layer and all hidden layers extract features from the sample input data by performing unsupervised pre-training on each RBM. The last hidden layer and the output layer form a neural network (NN for short), and the output feature vector of the RBM is used as the input vector of the above NN. The weight matrix W l is the hidden layer H l-1 and hidden layer H l The symmetric connection weights between them. As the basic unit of DBN, each RBM contains a visible layer and a hidden layer. The states of the visible layer and the hidden layer can be represented as v and h respectively. RBM is an energy model based on thermodynamics, and its energy function can be expressed as: Among them, w ij is the symmetric connection weight between the visual neurons and the hidden neurons; a i is the visual neuron v i Deviation of b j is the bias of the hidden neuron; n v is the number of neurons in the visual layer; n h is the number of neurons in the hidden layer. The joint probability distribution of the hidden layer and the visible layer can be defined as: When the state of the visible layer is known, h j The individual activation probability of can be derived by the following formula 1: Similarly, given h, v i The conditional probability can be calculated using the following formula 2:

[0046] In order to determine the parameters w, a, and b of the RBM, a fast learning algorithm called contrast divergence (CD) can be used to iteratively train a single RBM. The CD algorithm only needs one step of Gibbs sampling to obtain a good approximation. The learning process is as follows Figure 6 As shown in Figure 1, first, the sample input data is regarded as the state of the visual layer, and the state of the hidden layer is calculated by formula 1; then, the reconstructed state of the visual layer can be obtained by formula 2, and finally, the reconstructed state of the hidden layer can be obtained by formula 1. Therefore, the parameter update criterion of RBM can be expressed as: Among them, η is the learning rate of the CD algorithm; v′ is the state of the i-th neuron in the reconstructed visual layer; h′ is the state of the j-th neuron in the reconstructed hidden layer; <·> represents the mathematical expectation of the variable. As mentioned above, the parameters w, a, and b are initialized in the unsupervised pre-training process based on CD. Once the parameters of the first RBM are determined, the hidden layer output is regarded as the visual layer input of the second RBM. All RBMs of DBN update the corresponding network parameters layer by layer. On the other hand, the top layer of DBN adopts a feedforward neural network, and supervised training is performed based on error back propagation, while fine-tuning the parameters w, a, and b. In summary, the training process of DBN can be completed by combining the unsupervised pre-training process and the supervised fine-tuning process.

[0047] In order to further quickly establish the above-mentioned target back propagation neural network, the above-mentioned step S202 of the present application can be implemented by the following steps: step S2025, obtaining the initial back propagation neural network, and using a genetic algorithm to process the initial parameters of the above-mentioned back propagation neural network to obtain the target parameters, wherein the above-mentioned initial parameters include an initial threshold and an initial weight; step S2026, bringing the above-mentioned target parameters into the above-mentioned initial back propagation neural network to obtain the above-mentioned target back propagation neural network.

[0048] Specifically, the BP neural network is a multi-layer feedforward neural network that uses a back-propagation algorithm to train the weights of the network. The network is composed of many neurons working in parallel and has nonlinear adaptive dynamic characteristics of learning ability, memory function, computing ability and intelligent processing. The BP network consists of an input layer, multiple hidden layers and an output layer, and information is transmitted between nodes in each layer through unidirectional connections. The operation of the BP neural network involves two main stages: the forward flow of signals and the reverse transmission of errors. In the forward flow stage, the input data is converted through the hidden layer and then passed to the output layer. If the output error exceeds the set threshold, the network will start the error back-propagation mechanism to adjust the weights and thresholds to reduce the error. In the BP neural network, the initial weights and thresholds are randomly generated, and the selection of initial values ​​has a direct impact on the calculation results. Improper selection of initial values ​​can easily lead to slow convergence or fall into local minima and fail to converge. Therefore, the genetic algorithm is used here to obtain the optimal network weights and thresholds as the initial network weights and thresholds of the BP neural network model. The BP neural network optimized based on the genetic algorithm is recorded as the GA-BP model, and its calculation process is as follows Figure 7 shown.

[0049] The above step S202 can also be implemented in other ways, for example: step S2027, obtaining multiple historical impact data and multiple historical cost data, wherein the above historical impact data is the above cost impact data obtained before the current moment, and the above historical cost data is the above cost data obtained before the current moment; step S2028, inputting the above cost impact data into the above target back propagation neural network to obtain the calculation results corresponding to the above cost impact data, wherein the above target back propagation neural network is trained using multiple sets of training data, and each set of training data in the above multiple sets of training data includes the above historical impact data and the above historical cost data corresponding to the above historical impact data. This method can further quickly obtain the calculation results corresponding to the cost impact data.

[0050] Specifically, the above historical cost data can be obtained by collecting historical substation primary equipment maintenance project settlement data, and the above historical impact data can be obtained by analyzing the project cost composition and the key influencing factors that determine the total project cost.

[0051] In some embodiments, the above step S202 can be specifically implemented by the following steps: step S2029, obtaining a plurality of historical impact data and a plurality of historical cost data, wherein the above historical impact data is the above cost impact data obtained before the current moment, and the above historical cost data is the above cost data obtained before the current moment; step S20210, according to the formula Construct the above target grey prediction model, μ is the endogenous control grey number, a is the development grey number, k is a positive integer greater than or equal to 0, X (0) (1) is the first data set X (0) The first value of The second data set X (1) The k+1th value of , the first data group includes a plurality of the historical impact data arranged in time, and the second data group is generated by accumulating the first data group. This method can further quickly construct the target grey prediction model.

[0052] Specifically, grey prediction has the characteristics of simple modeling and small amount of data required. Grey model has relative advantages in dealing with uncertain systems with "small samples and little data", and can obtain high-precision prediction results even when data is relatively scarce. The modeling process is as follows:

[0053] First, a systematic correlation analysis method is proposed to conduct a quantitative comparative analysis of the system development trend. X (0) (k) = {X (0) (1), X (0) (2), ..., X (0) (n)}, then the correlation coefficient is defined as: in, is the kth point X (0) and The absolute error of is the minimum difference between the two levels; is the maximum difference between the two levels; ρ is called the resolution, 0<ρ<1, and ρ=0.5 is generally taken. Called X (0 )(k) and degree of correlation.

[0054] Second, establish the corresponding differential equation model, assuming that the time series X (0) There are n observations, denoted as X (0) ={X (0) (1),X (0) (2),...,X (0) (n)}, generate a new sequence X by accumulation (1) ={X (1) (1),X (1) (2),...,X (1) (n)}, then the corresponding differential equation of the GM(1,1) model is: Among them, a is the development gray number; μ is the endogenous control gray number. is the parameter vector to be estimated, The least square method can be used to solve it. The solution is: Solving the differential equation, we can get the prediction model:

[0055] Third, model testing. Grey prediction tests generally include residual test, correlation test and posterior difference test.

[0056] The above step S203 can also be implemented by the following steps: Step S2031, according to the formula: Calculate preliminary cost data in, is the predicted value of the above target deep belief network at the nth sample point, Back propagate the predicted value of the neural network at the nth sample point for the above target, is the predicted value of the target grey prediction model at the nth sample point, and the sample points correspond to the cost impact data one by one; step S2032, according to the formula and Calculate the loss function L, where q is the quantile, I is the indicator function, and y n is the actual cost data corresponding to the nth sample point, β q is the parameter vector of quantile q; Step S2033, determine the β corresponding to the case where the function value of the above loss function is the minimum q is the target parameter vector β q '; Step S2034, according to the formula and Calculate the upper limit value U n , where 1-α is the confidence level; Step S2035, according to the formula and Calculate the lower limit value L n ; Step S2036, calculate the above target cost data based on the above upper limit value and the above lower limit value.

[0057] Specifically, F-1(q|·) represents the conditional q quantile of the maintenance cost distribution; the calculation process of the above loss function is as follows:

[0058] The above step S2036 can also be implemented by the following steps: Step S20361, determine the preliminary interval according to the above upper limit value and the above lower limit value; Step S20362, according to the formula Calculate the credibility of the above-mentioned preparation interval PICP, where, when the above-mentioned preparation cost data corresponding to the above-mentioned nth sample point belongs to the above-mentioned preparation interval, c n is 1, and the above-mentioned preliminary cost data corresponding to the above-mentioned nth sample point does not belong to the above-mentioned preliminary interval, c nis 0, N is the total number of the above sample points; Step S20363, according to the formula Calculate the average bandwidth PINA W of the above-mentioned preparation interval; Step S20364, when the credibility of the above-mentioned preparation interval is greater than the credibility threshold and the average bandwidth of the above-mentioned preparation interval is greater than the width threshold, determine the above-mentioned preparation interval as the above-mentioned target cost data. This method can further improve the reliability of the above-mentioned target cost data.

[0059] Specifically, in addition to using the above-mentioned credibility and average bandwidth as evaluation interval indicators, other methods can also be used for evaluation.

[0060] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for determining substation maintenance cost data of the present application will be described in detail below in conjunction with specific embodiments.

[0061] This embodiment relates to a specific method for determining substation maintenance cost data, such as Figure 8 As shown, the following steps are included:

[0062] Step S1: Collecting settlement data of substation primary equipment maintenance project;

[0063] Step S2: Perform characteristic analysis on historical project data to obtain key influencing factors that determine the total project cost;

[0064] Step S3: cost estimation model based on deep belief network, cost estimation model based on genetic algorithm optimized BP neural network, cost estimation model based on grey prediction;

[0065] Step S4: combining the prediction results of different base models to establish a probability interval prediction model based on quantile regression average;

[0066] Step S5: Obtaining the interval prediction result of the project cost.

[0067] The embodiment of the present application also provides a device for determining substation maintenance cost data. It should be noted that the device for determining substation maintenance cost data in the embodiment of the present application can be used to execute the method for determining substation maintenance cost data provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0068] The following is an introduction to the device for determining substation maintenance cost data provided in an embodiment of the present application.

[0069] Fig. 9 Schematic diagram of a device for determining substation maintenance cost data according to an embodiment of the present application. Fig. 9 As shown, the device comprises:

[0070] The acquisition unit 10 is used to acquire cost impact data, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the first m positions, the cost data is the maintenance cost of the primary equipment of the substation, and the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation, and the process parameters of the substation, and m is a positive integer greater than 0;

[0071] Specifically, the substation primary equipment maintenance project consists of multiple processes, and the total cost of the project is the sum of all process costs. The specific cost data is composed of Figure 3 As shown. Due to the complex composition and wide range of process costs, a lot of business logic is required when using traditional methods to make cost budgets, especially the quota type, which requires detailed quota files for data such as the unit price of manpower, machine and materials. The workload is huge and there is a greater risk of error. Therefore, it is necessary to analyze the sample data to identify the key influencing factors of the project cost as the basic data support for establishing the cost estimation model. By sorting out and analyzing the historical settlement audit report data of the substation primary equipment maintenance project, it is concluded that the same type of projects usually have a large number of the same processes. In the substation primary equipment maintenance project, the project parameters and equipment parameters determine the fixed process and its project volume. Therefore, the project parameters, equipment parameters, variable process parameters, etc. can be the key influencing factors of the project cost.

[0072] A calculation unit 20 is used to construct a predetermined model, and use the predetermined model to calculate the cost impact data to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network and a target grey prediction model, and the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one;

[0073] Specifically, the structural diagram of the above predetermined model is as follows: Figure 4 As shown, by building a machine learning-based engineering cost estimation model framework, namely the above-mentioned predetermined model, an input and output sample set is established according to historical engineering data to obtain the fitting relationship between engineering parameters, equipment parameters, process parameters and engineering costs.

[0074] The processing unit 30 is used to perform fusion processing on the multiple calculation results to obtain target cost data, wherein the fusion processing at least includes quantile regression average processing.

[0075] Specifically, quantile regression is a statistical technique for estimating conditional quantiles. Unlike common mean regression, quantile regression can provide more comprehensive information about the conditional distribution of the response variable, not just the mean. In traditional linear regression, the focus is on how the expected value (or mean) of the dependent variable changes as the independent variable changes. However, in some cases, other quantiles (such as the median or quartiles) may also be of interest. Quantiles can estimate these conditional quantiles to obtain a deeper understanding of the data distribution. Quantile regression mean treatment refers to a special case of using quantile regression to estimate the conditional mean. In this case, a quantile of 0.5, that is, the median, can be selected for regression analysis. Since the median is insensitive to outliers, this method can provide a robust mean estimate.

[0076] Through this embodiment, a device for determining the maintenance cost data of a substation is provided, wherein an acquisition unit acquires cost impact data, the cost impact data is impact data that has an impact on the cost data and the impact degree is in the top m positions, the cost data is the maintenance cost of the primary equipment of the substation, and the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation, and the process parameters of the substation; the calculation unit cloud uses a predetermined model to calculate the cost impact data to obtain multiple calculation results, and the predetermined model includes a target deep confidence network, a target back propagation neural network, and a target gray prediction model; the processing unit performs fusion processing on the multiple calculation results to obtain the target cost data, and the fusion processing at least includes quantile regression average processing. By combining the prediction results of three different base models, namely the target deep confidence network, the target back propagation neural network, and the target gray prediction model, and fusing the three calculation results by quantile regression average, the maintenance cost of the primary equipment of the substation is predicted, so as to improve the reliability of the prediction results of the measurement model in engineering practice, and solve the problem of insufficient reliability in the process of determining cost data in the prior art.

[0077] In the specific implementation process, the computing unit includes a first acquisition module, a training module, a processing module and a second acquisition module. The first acquisition module is used to acquire an initial deep belief network, wherein the initial deep belief network includes a plurality of restricted Boltzmann machines and a back propagation neural network, wherein the plurality of restricted Boltzmann machines are used to extract the features of the input data of the initial deep belief network layer by layer, and the back propagation neural network is used for regression prediction; the training module is used to use an unsupervised learning algorithm to train the weights of the initial deep belief network and the bias of the initial deep belief network layer by layer to pre-train the initial deep belief network; the processing module is used to use forward propagation, back propagation and supervised learning algorithms to process the initial deep belief network to train the initial deep belief network; the second acquisition module is used to acquire the mean absolute percentage error and root mean square error of the initial deep belief network, and when the mean absolute percentage error is less than the first error threshold and the root mean square error is less than the second error threshold, the initial deep belief network is determined to be the target deep belief network. The device can further quickly establish the target deep belief network.

[0078] Specifically, the deep belief network (DBN) is an efficient unsupervised learning algorithm proposed by Hinton in 2006. It can be regarded as a combination of a series of restricted Boltzmann machines (RBM). Through the integration of deep learning and feature learning, DBN has good data fitting ability and fast analysis ability. The initial parameters of DBN can be obtained layer by layer through unsupervised pre-training, thereby effectively solving the related problems caused by the random initialization parameters of traditional neural networks. On the other hand, pre-training can effectively avoid overfitting and underfitting problems. The structure of the DBN prediction model is as follows: Figure 5 As shown in the figure, the DBN model consists of multiple RBMs and a single-layer back-propagation neural network (BP neural network for short). The BP neural network has 1 input layer, 1 hidden layer and 1 output layer. The input layer and all hidden layers extract features from the sample input data by performing unsupervised pre-training on each RBM. The last hidden layer and the output layer form a neural network (NN for short), and the output feature vector of the RBM is used as the input vector of the above NN. The weight matrix W l is the hidden layer H l-1 and hidden layer H lThe symmetric connection weights between them. As the basic unit of DBN, each RBM contains a visible layer and a hidden layer. The states of the visible layer and the hidden layer can be represented as v and h respectively. RBM is an energy model based on thermodynamics, and its energy function can be expressed as: Among them, w ij is the symmetric connection weight between the visual neurons and the hidden neurons; a i is the visual neuron v i Deviation of b j is the bias of the hidden neuron; n v is the number of neurons in the visual layer; n h is the number of neurons in the hidden layer. The joint probability distribution of the hidden layer and the visible layer can be defined as: When the state of the visible layer is known, h j The individual activation probability of can be derived by the following formula 1: Similarly, given h, v i The conditional probability can be calculated using the following formula 2:

[0079] In order to determine the parameters w, a, and b of the RBM, a fast learning algorithm called contrast divergence (CD) can be used to iteratively train a single RBM. The CD algorithm only needs one step of Gibbs sampling to obtain a good approximation. The learning process is as follows Figure 6 As shown in Figure 1, first, the sample input data is regarded as the state of the visual layer, and the state of the hidden layer is calculated by formula 1; then, the reconstructed state of the visual layer can be obtained by formula 2, and finally, the reconstructed state of the hidden layer can be obtained by formula 1. Therefore, the parameter update criterion of RBM can be expressed as: Among them, η is the learning rate of the CD algorithm; v′ is the state of the i-th neuron in the reconstructed visual layer; h′ is the state of the j-th neuron in the reconstructed hidden layer; <·> represents the mathematical expectation of the variable. As mentioned above, the parameters w, a, and b are initialized in the unsupervised pre-training process based on CD. Once the parameters of the first RBM are determined, the hidden layer output is regarded as the visual layer input of the second RBM. All RBMs of DBN update the corresponding network parameters layer by layer. On the other hand, the top layer of DBN adopts a feedforward neural network, and supervised training is performed based on error back propagation, while fine-tuning the parameters w, a, and b. In summary, the training process of DBN can be completed by combining the unsupervised pre-training process and the supervised fine-tuning process.

[0080] In order to further quickly establish the above-mentioned target back-propagation neural network, the above-mentioned computing unit of the present application also includes a third acquisition module and an input module, wherein the third acquisition module is used to obtain the initial back-propagation neural network, and use a genetic algorithm to process the initial parameters of the above-mentioned back-propagation neural network to obtain the target parameters, wherein the above-mentioned initial parameters include an initial threshold and an initial weight; the input module is used to bring the above-mentioned target parameters into the above-mentioned initial back-propagation neural network to obtain the above-mentioned target back-propagation neural network.

[0081] Specifically, the BP neural network is a multi-layer feedforward neural network that uses a back-propagation algorithm to train the weights of the network. The network is composed of many neurons working in parallel and has nonlinear adaptive dynamic characteristics of learning ability, memory function, computing ability and intelligent processing. The BP network consists of an input layer, multiple hidden layers and an output layer, and information is transmitted between nodes in each layer through unidirectional connections. The operation of the BP neural network involves two main stages: the forward flow of signals and the reverse transmission of errors. In the forward flow stage, the input data is converted through the hidden layer and then passed to the output layer. If the output error exceeds the set threshold, the network will start the error back-propagation mechanism to adjust the weights and thresholds to reduce the error. In the BP neural network, the initial weights and thresholds are randomly generated, and the selection of initial values ​​has a direct impact on the calculation results. Improper selection of initial values ​​can easily lead to slow convergence or fall into local minima and fail to converge. Therefore, the genetic algorithm is used here to obtain the optimal network weights and thresholds as the initial network weights and thresholds of the BP neural network model. The BP neural network optimized based on the genetic algorithm is recorded as the GA-BP model, and its calculation process is as follows Figure 7 shown.

[0082] The calculation unit includes a fourth acquisition module and an input module, wherein the fourth acquisition module is used to acquire multiple historical impact data and multiple historical cost data, wherein the historical impact data is the cost impact data acquired before the current moment, and the historical cost data is the cost data acquired before the current moment; the input module is used to input the cost impact data into the target back propagation neural network to obtain the calculation results corresponding to the cost impact data, wherein the target back propagation neural network is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes the historical impact data and the historical cost data corresponding to the historical impact data. The device can further quickly acquire the calculation results corresponding to the cost impact data.

[0083] Specifically, the above historical cost data can be obtained by collecting historical substation primary equipment maintenance project settlement data, and the above historical impact data can be obtained by analyzing the project cost composition and the key influencing factors that determine the total project cost.

[0084] In some embodiments, the calculation unit further includes a fourth acquisition module and a construction module, wherein the fourth acquisition module is used to acquire a plurality of historical impact data and a plurality of historical cost data, wherein the historical impact data is the cost impact data acquired before the current moment, and the historical cost data is the cost data acquired before the current moment; the construction module is used to calculate the cost impact data according to the formula Construct the above target grey prediction model, μ is the endogenous control grey number, a is the development grey number, k is a positive integer greater than or equal to 0, X (0) (1) is the first data set X (0) The first value of The second data set X (1) The k+1th value of the first data group includes a plurality of the historical impact data arranged in time, and the second data group is generated by accumulating the first data group. The device can further quickly construct the target grey prediction model.

[0085] Specifically, grey prediction has the characteristics of simple modeling and small amount of data required. Grey model has relative advantages in dealing with uncertain systems with "small samples and little data", and can obtain high-precision prediction results even when data is relatively scarce. The modeling process is as follows:

[0086] First, a systematic correlation analysis method is proposed to conduct a quantitative comparative analysis of the system development trend. X (0) (k) = {X (0) (1), X (0) (2), ..., X (0) (n)}, then the correlation coefficient is defined as: in, is the kth point X (0) and The absolute error of is the minimum difference between the two levels; is the maximum difference between the two levels; ρ is called the resolution, 0<ρ<1, and ρ=0.5 is generally taken. Called X (0) (k) degree of correlation.

[0087] Second, establish the corresponding differential equation model, assuming that the time series X (0) There are n observations, denoted as X (0) ={X (0) (1),X (0) (2),...,X (0) (n)}, generate a new sequence X by accumulation (1) ={X(1) (1),X (1) (2),...,X (1) (n)}, then the corresponding differential equation of the GM(1,1) model is: Among them, a is the development gray number; μ is the endogenous control gray number. is the parameter vector to be estimated, The least square method can be used to solve it. The solution is: Solving the differential equation, we can get the prediction model:

[0088] Third, model testing. Grey prediction tests generally include residual test, correlation test and posterior difference test.

[0089] The processing unit includes a first calculation module, a second calculation module, a determination module, a third calculation module, a fourth calculation module and a fifth calculation module, wherein the first calculation module is used to determine the value of the first calculation module according to the formula: Calculate preliminary cost data in, is the predicted value of the above target deep belief network at the nth sample point, Back propagate the predicted value of the neural network at the nth sample point for the above target, is the predicted value of the target grey prediction model at the nth sample point, and the sample points correspond to the cost impact data one by one; the second calculation module is used to calculate the value according to the formula: and Calculate the loss function L, where q is the quantile, I is the indicator function, and y n is the actual cost data corresponding to the nth sample point, β q is the parameter vector of the quantile q; the determination module is used to determine the corresponding β when the function value of the above loss function is the minimum q is the target parameter vector β q '; The third calculation module is used according to the formula and Calculate the upper limit value U n , where 1-α is the confidence level; the fourth calculation module is used to calculate the confidence level according to the formula and Calculate the lower limit value L n ; The fifth calculation module is used to calculate the above-mentioned target cost data based on the above-mentioned upper limit value and the above-mentioned lower limit value.

[0090] Specifically, F-1(q|·) represents the conditional q quantile of the maintenance cost distribution; the calculation process of the above loss function is as follows:

[0091] The fifth calculation module includes a determination submodule, a first calculation submodule and a second calculation submodule, wherein the determination submodule is used to determine the preliminary interval according to the upper limit value and the lower limit value; the first calculation submodule is used to determine the preliminary interval according to the formula Calculate the credibility of the above-mentioned preparation interval PICP, where, when the above-mentioned preparation cost data corresponding to the above-mentioned nth sample point belongs to the above-mentioned preparation interval, c n is 1, and the above-mentioned preliminary cost data corresponding to the above-mentioned nth sample point does not belong to the above-mentioned preliminary interval, c n is 0, N is the total number of the above sample points; the second calculation submodule is used according to the formula: Calculate the average bandwidth PINAW of the above-mentioned preparation interval; Step S20364, when the credibility of the above-mentioned preparation interval is greater than the credibility threshold and the average bandwidth of the above-mentioned preparation interval is greater than the width threshold, determine the above-mentioned preparation interval as the above-mentioned target cost data. This device can further improve the reliability of the above-mentioned target cost data.

[0092] Specifically, in addition to using the above-mentioned credibility and average bandwidth as evaluation interval indicators, other methods can also be used for evaluation.

[0093] The above-mentioned device for determining the maintenance cost data of a substation includes a processor and a memory. The above-mentioned acquisition unit, calculation unit, and processing unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0094] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the substation maintenance cost data can be determined by adjusting kernel parameters.

[0095] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0096] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for determining the substation maintenance cost data.

[0097] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the method for determining the substation maintenance cost data is executed when the program is run.

[0098] An embodiment of the present invention provides a device, the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the steps of the method for determining substation maintenance cost data are implemented. The device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0099] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes at least the steps of a method for determining substation maintenance cost data.

[0100] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0101] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0104] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0105] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0106] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0107] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0108] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for determining substation maintenance cost data, characterized in that: include: Acquire cost impact data, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the first m positions, the cost data is the maintenance cost of the primary equipment of the substation, the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation and the process parameters of the substation, and m is a positive integer greater than 0; Constructing a predetermined model, and using the predetermined model to calculate the cost impact data to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network, and a target grey prediction model, the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one; A plurality of the calculation results are fused to obtain target cost data, wherein the fusion processing at least includes quantile regression averaging processing.

2. The method according to claim 1, characterized in that Constructing the target deep belief network, including: Acquire an initial deep belief network, wherein the initial deep belief network includes a plurality of restricted Boltzmann machines and a back propagation neural network, the plurality of restricted Boltzmann machines are used to extract features of input data of the initial deep belief network layer by layer, and the back propagation neural network is used for regression prediction; Using an unsupervised learning algorithm to train the weights of the initial deep belief network and the bias of the initial deep belief network layer by layer to pre-train the initial deep belief network; Processing the initial deep belief network using forward propagation, back propagation, and a supervised learning algorithm to train the initial deep belief network; The mean absolute percentage error and the root mean square error of the initial deep belief network are obtained, and when the mean absolute percentage error is less than a first error threshold and the root mean square error is less than a second error threshold, the initial deep belief network is determined to be the target deep belief network.

3. The method according to claim 1, characterized in that Constructing the target back propagation neural network, including: Acquire an initial back propagation neural network, and process the initial parameters of the back propagation neural network using a genetic algorithm to obtain target parameters, wherein the initial parameters include an initial threshold and an initial weight; The target parameters are introduced into the initial back-propagation neural network to obtain the target back-propagation neural network.

4. The method according to claim 1, characterized in that The target back propagation neural network is used to calculate the cost impact data to obtain multiple calculation results, including: Acquire a plurality of historical impact data and a plurality of historical cost data, wherein the historical impact data is the cost impact data acquired before the current moment, and the historical cost data is the cost data acquired before the current moment; The cost impact data is input into the target back propagation neural network to obtain a calculation result corresponding to the cost impact data, wherein the target back propagation neural network is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes the historical impact data and the historical cost data corresponding to the historical impact data.

5. The method according to claim 1, characterized in that: Constructing the target grey prediction model includes: Acquire a plurality of historical impact data and a plurality of historical cost data, wherein the historical impact data is the cost impact data acquired before the current moment, and the historical cost data is the cost data acquired before the current moment; According to the formula Construct the target grey prediction model, μ is the endogenous control grey number, a is the development grey number, k is a positive integer greater than or equal to 0, X (0) (1) is the first data set X (0) The first value of The second data set X (1) The first data group includes a plurality of the historical impact data arranged in time, and the second data group is generated by accumulating the first data groups.

6. The method according to claim 1, characterized in that The multiple calculation results are fused to obtain target cost data, including: According to the formula Calculate preliminary cost data in, is the predicted value of the target deep belief network at the nth sample point, Back propagate the predicted value of the target neural network at the nth sample point, is the predicted value of the target grey prediction model at the nth sample point, and the sample points correspond to the cost influencing data one by one; According to the formula and Calculate the loss function L, where q is the quantile, I is the indicator function, and y n is the actual cost data corresponding to the nth sample point, β q is the parameter vector of quantile q; Determine the β corresponding to the minimum value of the loss function q is the target parameter vector β q ′; According to the formula and Calculate the upper limit value U n , where 1-α is the confidence level; According to the formula and Calculate the lower limit value L n ; The target cost data is calculated according to the upper limit value and the lower limit value.

7. The method according to claim 6, characterized in that Calculating the target cost data according to the upper limit value and the lower limit value includes: Determining a preliminary interval according to the upper limit value and the lower limit value; According to the formula Calculate the credibility of the preparatory interval PICP, wherein, when the preparatory cost data corresponding to the nth sample point belongs to the preparatory interval, c n is 1, and the preliminary cost data corresponding to the nth sample point does not belong to the preliminary interval, c n is 0, and N is the total number of sample points; According to the formula Calculate the average bandwidth PINAW of the preparation interval; When the credibility of the preliminary interval is greater than a credibility threshold and the average bandwidth of the preliminary interval is greater than a bandwidth threshold, the preliminary interval is determined to be the target cost data.

8. A device for determining substation maintenance cost data, characterized in that: include: an acquisition unit, configured to acquire cost impact data, wherein the cost impact data is impact data that has an impact on the cost data and the impact degree is in the first m positions, the cost data is the maintenance cost of the primary equipment of the substation, the cost impact data includes at least one of the engineering parameters of the substation, the equipment parameters of the substation and the process parameters of the substation, and m is a positive integer greater than 0; A calculation unit, used to construct a predetermined model, and use the predetermined model to calculate the cost impact data to obtain multiple calculation results, wherein the predetermined model includes a target deep belief network, a target back propagation neural network and a target grey prediction model, the cost impact data and the predetermined model correspond one-to-one, and the cost impact data and the calculation results correspond one-to-one; A processing unit is used to perform fusion processing on the multiple calculation results to obtain target cost data, wherein the fusion processing at least includes quantile regression average processing.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.