Method and device for determining theoretical line loss of power distribution network based on deep belief network

Through the method based on deep confidence network, the line loss determination model is constructed and trained, which solves the problems of low accuracy and slow speed in the theoretical line loss calculation of the power grid, and achieves fast, efficient and accurate calculations.

CN118964895APending Publication Date: 2024-11-15GUANGDONG POWER GRID CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411144011.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the prior art, it is difficult to calculate quickly and efficiently accurately when calculating the theoretical line loss of the power grid, and the traditional methods have low calculation accuracy and slow speed, which cannot meet the actual production needs.

Method used

Using a deep confidence network-based method, a line loss determination model is constructed and trained by obtaining and standardizing line loss feature parameters, and multiple sets of training data are used for iterative training and reverse fine-tuning until the convergence conditions are reached, and an accurate theoretical line loss is obtained.

Benefits of technology

It realizes the rapid and efficient and accurate calculation of theoretical line loss, improves the calculation speed and accuracy, and solves the problems of low and slow calculation accuracy in traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964895A_ABST
    Figure CN118964895A_ABST
Patent Text Reader

Abstract

The invention discloses a method and a device for determining theoretical line loss of a power distribution network based on a deep belief network. The method comprises the following steps: acquiring a line loss characteristic parameter set of the power distribution network; the standardized line loss characteristic parameter set is input into a line loss determination model, the line loss determination model is used for processing the line loss characteristic parameters, the theoretical line loss of the power distribution network is obtained, the line loss determination model is obtained by using multiple sets of training data to train in a deep belief network mode, and the theoretical line loss of the power distribution network is obtained. Each group in the multiple groups of training data comprises a sample line loss characteristic parameter set and sample theoretical line loss corresponding to the sample line loss characteristic parameter set. The technical problem that the theoretical line loss is difficult to quickly, efficiently and accurately calculate when the theoretical line loss of the power grid is calculated by using a traditional method in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power grid line loss calculation, and in particular to a method and device for determining theoretical line loss of a distribution network based on a deep belief network. Background Art

[0002] Theoretical line loss is an important economic indicator of power supply enterprises. It reflects the planning, design, production, operation and management level of power system. Therefore, power supply companies at all levels must calculate, count, analyze and evaluate line loss. Accurate and fast calculation of line loss helps to reasonably reduce energy loss, save resources, formulate loss reduction plans and improve economic benefits.

[0003] At present, there are many calculation methods for theoretical line loss of power grid, which can basically be summarized into three categories: the first is to use traditional calculation methods, such as equivalent resistance method, root mean square current method, average current method, power flow method, etc. Among them, the calculation accuracy of power flow method is closest to the actual theoretical line loss. In actual production work, the data integrity and validity often cannot meet the calculation requirements of power flow method, and algorithms with lower calculation accuracy are usually used to calculate theoretical line loss; the second is to improve the calculation method based on traditional methods with less data requirements and lower calculation accuracy, and improve the accuracy and practicality of line loss calculation by optimizing the calculation method of load curve shape coefficient, introducing three-phase state characteristics, optimizing line loss calculation interval, etc.; the third is to use intelligent algorithms for line loss calculation, such as clustering algorithm, ant colony algorithm, traditional neural network algorithm, radial basis function neural network algorithm and other neural network improvement algorithms.

[0004] Among the traditional calculation methods, equivalent algorithms such as the equivalent resistance method, root mean square current method, and average current method have low calculation accuracy, and the power flow method is often unable to calculate in actual production work due to the limitations of data integrity and validity, and the calculation speed is slow; the improved calculation method based on the traditional method with less data requirements and lower calculation accuracy requires the collection of more additional operating data or an increase in the number of calculations, which increases the difficulty of data collection and the complexity of calculations; the use of intelligent algorithms for line loss calculations is mostly based on unsupervised calculation models, with low calculation accuracy and low convergence speed.

[0005] With regard to the problem in the above-mentioned related technologies that it is difficult to quickly and efficiently calculate the theoretical line loss accurately when using traditional methods to calculate the theoretical line loss of the power grid, no effective solution has been proposed so far. Summary of the invention

[0006] The embodiments of the present invention provide a method and device for determining the theoretical line loss of a distribution network based on a deep belief network, so as to at least solve the technical problem in the related art that it is difficult to quickly and efficiently calculate the theoretical line loss accurately when using traditional methods to calculate the theoretical line loss of the power grid.

[0007] According to one aspect of an embodiment of the present invention, a method for determining a theoretical line loss of a distribution network based on a deep belief network is provided, comprising: obtaining a line loss feature parameter set of the distribution network, wherein the line loss feature parameter set is a collection of multiple line loss feature parameters, and each of the multiple line loss feature parameters is used to represent the loss characteristics of the distribution network; standardizing the multiple line loss feature parameters in the line loss feature parameter set; inputting the standardized line loss feature parameter set into a line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple groups of training data by means of a deep belief network, and each of the multiple groups of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0008] Optionally, performing standardization processing on the plurality of line loss characteristic parameters in the line loss characteristic parameter set includes: calculating an average value of each line loss characteristic parameter in the line loss characteristic parameter set according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any of the line loss characteristic parameter sets, n represents the total number of the line loss characteristic parameter sets, x ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; and calculates the standard deviation of each line loss characteristic parameter in the line loss characteristic parameter set according to a second formula, wherein the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; and determines the standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter according to a third formula, wherein the third formula is: X ij Represents the standard line loss characteristic parameter corresponding to the jth line loss characteristic parameter in the i-th line loss characteristic parameter set.

[0009] Optionally, before the standardized line loss characteristic parameter set is input into the line loss determination model to process the line loss characteristic parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, the distribution network theoretical line loss determination method based on the deep belief network also includes: obtaining multiple groups of sample line loss characteristic parameter sets, wherein each group of sample line loss characteristic parameter sets in the multiple groups of sample line loss characteristic parameter sets includes at least the following sample line loss characteristic parameters: sample line length, sample line model, sample distribution quantity, sample distribution model and sample distribution quantity; using Z score The standardization method standardizes multiple sample line loss feature parameters in multiple groups of sample line loss feature parameter sets; constructs an initial line loss determination model based on the deep belief network; uses multiple groups of sample line loss feature parameter sets to iteratively train the initial line loss determination model until the convergence condition is reached, and determines that the current initial line loss determination model is the line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches a preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than a deviation threshold.

[0010] Optionally, the initial line loss determination model is iteratively trained using multiple sets of the sample line loss feature parameter sets until a convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, including: using multiple sets of the sample line loss feature parameter sets to pre-train the initial line loss determination model, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters, and the RBM weight parameters of each neural layer are obtained, wherein the RBM weight parameters are the weight parameters of the restricted Boltzmann machine in the neural layer ; Use multiple sets of labeled sample line loss feature parameter sets to reversely fine-tune the initial line loss determination model after the pre-training to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and determine that the current initial line loss determination model is the line loss determination model, wherein the labeled sample line loss feature parameter set is the sample line loss feature parameter set with the corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0011] Optionally, multiple sets of labeled sample line loss feature parameter sets are used to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, including: a first acquisition step, inputting the labeled sample line loss feature parameter set into the pre-trained initial line loss determination model for processing, and obtaining a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is a predicted line loss obtained by the initial line loss determination model predicting the labeled sample line loss feature parameter set Theoretical line loss; a second acquisition step, acquiring a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; a third acquisition step, comparing the first deviation value with the deviation threshold to obtain a first comparison result; an optimization step, when the first comparison result indicates that the first deviation value is greater than the deviation threshold, optimizing the reverse fine-tuning process of the initial line loss determination model using an adaptive motion estimation algorithm; repeating the first acquisition step, the second acquisition step, the third acquisition step and the optimization step until the first deviation value is not greater than the deviation threshold, determining that the current initial line loss determination model is the line loss determination model.

[0012] Optionally, a plurality of sets of labeled sample line loss feature parameter sets are used to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, including: in the process of reverse fine-tuning the pre-trained initial line loss determination model using a plurality of sets of labeled sample line loss feature parameter sets, counting the number of iterations of the labeled sample line loss feature parameter sets; when the number of iterations reaches a preset number of iterations, determining the current initial line loss determination model to be the line loss determination model.

[0013] Optionally, after iteratively training the initial line loss determination model using multiple groups of the sample line loss feature parameter sets until convergence conditions are reached and determining that the current initial line loss determination model is the line loss determination model, the distribution network theoretical line loss determination method based on a deep belief network also includes: obtaining a test line loss feature parameter set and a test theoretical line loss corresponding to the test line loss feature parameter set; inputting the test line loss feature parameter set into the line loss determination model for processing to obtain a second predicted theoretical line loss corresponding to the test line loss feature parameter set, wherein the second predicted theoretical line loss is the predicted theoretical line loss obtained by the line loss determination model based on the test line loss feature parameter set; comparing a second deviation value between the second predicted theoretical line loss and the test theoretical line loss with the deviation threshold to obtain a second comparison result; when the second comparison result indicates that the second deviation value is less than the deviation threshold, it is determined that the line loss determination model has successfully passed the accuracy verification.

[0014] According to another aspect of an embodiment of the present invention, there is also provided a device for determining theoretical line loss of a distribution network based on a deep belief network, comprising: a first acquisition unit, used to acquire a line loss feature parameter set of the distribution network, wherein the line loss feature parameter set is a collection of multiple line loss feature parameters, and each of the multiple line loss feature parameters is used to represent the loss characteristics of the distribution network; a first processing unit, used to standardize the multiple line loss feature parameters in the line loss feature parameter set; a second acquisition unit, used to input the standardized line loss feature parameter set into a line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple groups of training data by means of a deep belief network, and each of the multiple groups of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0015] Optionally, the first processing unit includes: a first calculation module, configured to calculate an average value of each of the line loss characteristic parameters in the line loss characteristic parameter set according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any of the line loss characteristic parameter sets, n represents the total number of the line loss characteristic parameter sets, x ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; a second calculation module, used to calculate the standard deviation of each line loss characteristic parameter in the line loss characteristic parameter set according to a second formula, wherein the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; a first determination module, used to determine the standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter according to a third formula, wherein the third formula is: X ij Represents the standard line loss characteristic parameter corresponding to the jth line loss characteristic parameter in the i-th line loss characteristic parameter set.

[0016] Optionally, the distribution network theoretical line loss determination device based on deep belief network also includes: a third acquisition unit, which is used to input the line loss feature parameter set after standardization into the line loss determination model, so as to use the line loss determination model to process the line loss feature parameters to obtain the theoretical line loss of the distribution network, and then obtain multiple groups of sample line loss feature parameter sets, wherein each group of sample line loss feature parameter sets in the multiple groups of sample line loss feature parameter sets includes at least the following sample line loss feature parameters: sample line length, sample line model, sample distribution quantity, sample distribution model and sample distribution power; a second processing unit, which is used to adopt Z-division to obtain the sample line loss feature parameter set. A number normalization method is used to standardize multiple sample line loss feature parameters in multiple groups of sample line loss feature parameter sets; a construction unit is used to construct an initial line loss determination model based on the deep belief network; a first determination unit is used to iteratively train the initial line loss determination model using multiple groups of sample line loss feature parameter sets until a convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches a preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than a deviation threshold.

[0017] Optionally, the determination unit includes: an acquisition module, used to pre-train the initial line loss determination model using multiple groups of the sample line loss feature parameter sets, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters, and the RBM weight parameters of each neural layer are obtained, wherein the RBM weight parameters are the weight parameters of the restricted Boltzmann machine in the neural layer; a second determination module, used to reversely fine-tune the pre-trained initial line loss determination model using multiple groups of labeled sample line loss feature parameter sets, so as to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, wherein the labeled sample line loss feature parameter set is the sample line loss feature parameter set with the corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0018] Optionally, the second determination module includes: a first acquisition submodule, which is used to input the labeled sample line loss feature parameter set into the initial line loss determination model after the pre-training for processing, and obtain a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is the predicted theoretical line loss obtained by the initial line loss determination model predicting the labeled sample line loss feature parameter set; a second acquisition submodule, which is used to obtain a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; a third acquisition submodule, which is used to compare the first deviation value with the deviation threshold to obtain a first comparison result; an optimization submodule, which is used to optimize the reverse fine-tuning process of the initial line loss determination model using an adaptive motion estimation algorithm when the first comparison result indicates that the first deviation value is greater than the deviation threshold; a first determination submodule, which is used to repeatedly execute the first acquisition submodule, the second acquisition submodule, the third acquisition submodule and the optimization submodule until the first deviation value is not greater than the deviation threshold, and the current initial line loss determination model is determined to be the line loss determination model.

[0019] Optionally, the second determination module includes: a statistical submodule, used to count the number of iterations of the labeled sample line loss feature parameter set in the process of reverse fine-tuning the pre-trained initial line loss determination model using multiple sets of labeled sample line loss feature parameter sets; and a second determination submodule, used to determine that the current initial line loss determination model is the line loss determination model when the number of iterations reaches a preset number of iterations.

[0020] Optionally, the distribution network theoretical line loss determination device based on a deep belief network also includes: a fourth acquisition unit, which is used to iteratively train the initial line loss determination model using multiple groups of the sample line loss feature parameter sets until the convergence condition is reached, and then determine that the current initial line loss determination model is the line loss determination model, and then obtain a test line loss feature parameter set and a test theoretical line loss corresponding to the test line loss feature parameter set; a fifth acquisition unit, which is used to input the test line loss feature parameter set into the line loss determination model for processing to obtain a second predicted theoretical line loss corresponding to the test line loss feature parameter set, wherein the second predicted theoretical line loss is the predicted theoretical line loss obtained by the line loss determination model according to the test line loss feature parameter set; a sixth acquisition unit, which is used to compare the second deviation value between the second predicted theoretical line loss and the test theoretical line loss with the deviation threshold to obtain a second comparison result; and a second determination unit, which is used to determine that the line loss determination model has successfully passed the accuracy verification when the second comparison result indicates that the second deviation value is less than the deviation threshold.

[0021] According to another aspect of an embodiment of the present invention, a distribution network theoretical line loss determination system based on a deep belief network is also provided. The distribution network theoretical line loss determination system based on a deep belief network uses any of the above-mentioned distribution network theoretical line loss determination methods based on a deep belief network.

[0022] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided, wherein the computer-readable storage medium includes a stored program, wherein the program executes any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network.

[0023] According to another aspect of an embodiment of the present invention, a processor is further provided, wherein the processor is used to run a program, wherein when the program is running, any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network is executed.

[0024] According to another aspect of an embodiment of the present invention, a computer program product is also provided, comprising computer instructions, which, when executed by a processor, execute any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network.

[0025] In an embodiment of the present invention, a line loss characteristic parameter set of a distribution network is obtained, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; the multiple line loss characteristic parameters in the line loss characteristic parameter set are standardized; the standardized line loss characteristic parameter set is input into a line loss determination model, so as to process the line loss characteristic parameters using the line loss determination model to obtain a theoretical line loss of the distribution network, wherein the line loss determination model is trained by using multiple groups of training data through a deep belief network, and each of the multiple groups of training data includes: a sample line loss characteristic parameter set, and a sample theoretical line loss corresponding to the sample line loss characteristic parameter set. Through the above technical scheme, multiple characteristic parameters of the distribution network are selected as input variables of the line loss determination model, and then the input variables are processed by the line loss determination model based on deep belief network training to output the theoretical line loss value. The line loss determination model constructed by the deep belief network analyzes the correlation between the distribution network line loss characteristic value and the theoretical line loss, so as to quickly, efficiently and accurately calculate the theoretical line loss value. The calculation speed and accuracy of the theoretical line loss are improved, and the technical problem that it is difficult to quickly, efficiently and accurately calculate the theoretical line loss in the related technology when using traditional methods to calculate the theoretical line loss of the power grid is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 It is a hardware structure block diagram of a mobile terminal of a method for determining theoretical line loss of a distribution network based on a deep belief network according to an embodiment of the present invention;

[0028] Figure 2 is a flow chart of a method for determining theoretical line loss of a distribution network based on a deep belief network according to an embodiment of the present invention;

[0029] Figure 3 is a flowchart of training a linear deterministic model according to an embodiment of the present invention;

[0030] Figure 4 is a schematic diagram of a training linear determination model according to an embodiment of the present invention;

[0031] Figure 5 is a schematic diagram comparing the line loss calculation results of DBN and ANN according to an embodiment of the present invention;

[0032] Figure 6 is a schematic diagram of a device for determining theoretical line loss of a distribution network based on a deep belief network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] As introduced in the background technology, it is difficult to quickly and efficiently calculate the theoretical line loss accurately when using traditional methods to calculate the theoretical line loss of the power grid in the related art. In view of the above defects, a method and device for determining the theoretical line loss of a distribution network based on a deep belief network are provided in an embodiment of the present invention.

[0036] 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.

[0037] The method embodiments provided in the embodiments of the present invention 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 of a method for determining theoretical line loss of a distribution network based on a deep belief network 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 1The 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 1 Different configurations shown.

[0038] 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 theoretical line loss of a distribution network based on a deep belief network in an 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, to implement the above method. 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 instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof. The transmission device 106 is used to receive or send data via a network. The above network specific examples 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 Cont roller, 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.

[0039] According to an embodiment of the present invention, a method embodiment of a method for determining theoretical line loss in a distribution network based on a deep belief network 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 may be executed in an order different from that shown here.

[0040] Figure 2 is a flow chart of a method for determining theoretical line loss of a distribution network based on a deep belief network according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0041] Step S202, obtaining a line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network.

[0042] In this embodiment, the grid structure of the distribution network is complex and the operation conditions are diverse. The theoretical line loss flow calculation involves multi-dimensional data such as distribution network archives, topology, and operation. The required data is of various types and depends on the data quality, so the calculation speed is slow. The theoretical line loss of the distribution network mainly consists of line loss and distribution transformer loss. Considering the influence of these data indicators on line loss and the difficulty of obtaining them, line length, line model, number of distribution transformers, distribution transformer model, and distribution transformer power are selected as characteristic indicators of the distribution network (i.e., line loss characteristic parameters), i.e., the input of the line loss determination model.

[0043] In addition, a collection of a set of data of these characteristic indicators can be used as a line loss characteristic parameter set. When using the line loss determination model to calculate the theoretical line loss, one or more line loss characteristic parameter sets (that is, one or more sets of line loss characteristic parameters) can be input into the line loss determination model for processing to obtain the theoretical line loss corresponding to each line loss characteristic parameter set.

[0044] The following describes these characteristic indicators in detail: 1) Line length: for a certain line model, the longer the line length, the greater the resistance value. In this embodiment, the length of the main line and the length of the branch line of the medium-voltage line can be used as the characteristic value; 2) Line model: different line types usually have very different resistance values. In this embodiment, the total line resistance corresponding to each line model of the medium-voltage line can be used as the characteristic value; 3) Number of distribution transformers: the more transformers of each type, the greater the inherent loss of the distribution network. In this embodiment, the number of distribution transformers of different models can be used as the characteristic value; 4) Distribution transformer model: different line models have very different no-load loss and short-circuit loss values. In this embodiment, the total no-load loss and total short-circuit loss values ​​corresponding to the distribution transformer model can be used as the characteristic value; 5) Distribution transformer power: load directly affects line loss. For a certain distribution network scale, the greater the load of the distribution transformer under the line, the greater the line loss and transformer copper loss. In this embodiment, the total daily power of the distribution transformer connected to the medium-voltage line can be used as the characteristic value.

[0045] Step S204, performing standardization processing on multiple line loss characteristic parameters in the line loss characteristic parameter set.

[0046] In this embodiment, in order to reduce the impact of numerical and dimensional issues on the calculation, a Z-Score standardization method may be used to standardize multiple line loss characteristic parameters in a line loss characteristic parameter set to be input into the line loss determination model.

[0047] According to the above embodiment of the present invention, in the above step S204, the plurality of line loss characteristic parameters in the line loss characteristic parameter set are standardized, including: calculating the average value of each line loss characteristic parameter in the line loss characteristic parameter set according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any line loss characteristic parameter set, n represents the total number of line loss characteristic parameter sets, x represents the number of the line loss characteristic parameter sets, ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; the standard deviation of each line loss characteristic parameter in the line loss characteristic parameter set is calculated according to the second formula, where the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; and determines the standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter according to the third formula, wherein the third formula is: X ij Represents the standard line loss characteristic parameter corresponding to the j-th line loss characteristic parameter in the i-th line loss characteristic parameter set.

[0048] Specifically, assuming that there are i line loss characteristic parameter sets, each of which contains j line loss characteristic parameters, the formula can be used first: Calculate the average value of each line loss characteristic parameter in the i line loss characteristic parameter sets, where n represents the total number of line loss characteristic parameter sets, x ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, Represents the average value of the jth line loss characteristic parameter; then use the formula: Calculate the standard deviation of each line loss characteristic parameter, where s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; finally, the formula is used: Determine for each line loss characteristic parameter x ij Standard line loss characteristic parameter X after standardization ij .

[0049] It should be noted that the sorting order of the line loss characteristic parameters in each line loss characteristic parameter set is not specifically limited in the embodiment of the present invention, but the sorting order of the line loss characteristic parameters in each line loss characteristic parameter set should be consistent.

[0050] Step S206, inputting the standardized line loss feature parameter set into the line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple sets of training data through a deep belief network, and each of the multiple sets of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0051] In this embodiment, the trained line loss determination model may be used to process the input line loss feature parameter set to output the corresponding theoretical line loss.

[0052] According to the above embodiment of the present invention, before the above step S206, that is, before the standardized line loss feature parameter set is input into the line loss determination model to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, the distribution network theoretical line loss determination method based on the deep belief network also includes: obtaining multiple groups of sample line loss feature parameter sets, wherein each group of sample line loss feature parameter sets in the multiple groups of sample line loss feature parameter sets includes at least the following sample line loss feature parameters: sample line length, sample line model, sample distribution quantity, sample distribution model and sample power distribution capacity; use Z-score normalization method to standardize multiple sample line loss feature parameters in multiple groups of sample line loss feature parameter sets; build an initial line loss determination model based on deep belief network; use multiple groups of sample line loss feature parameter sets to iteratively train the initial line loss determination model until the convergence condition is reached, and determine the current initial line loss determination model as the line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches the preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than the deviation threshold.

[0053] Combine the following Figure 3 and Figure 4 The construction and training process of the line loss determination model in the above embodiment of the present invention are described in detail. Figure 3 is a flowchart of training a linear deterministic model according to an embodiment of the present invention; Figure 4 is a schematic diagram of a training linear deterministic model according to an embodiment of the present invention.

[0054] like Figure 3 As shown, multiple groups of sample line loss feature parameter sets can be obtained, and part of them can be used as training data to train the line loss determination model, and the other part can be used as test data to verify the initially trained line loss determination model to verify its accuracy.

[0055] In addition, when training the line loss determination model, it is also necessary to use the Z-Score standardization method (Z score standardization method) to standardize each sample line loss feature parameter in the input sample line loss feature parameter set, so that the trained line loss determination model can correctly understand and predict the newly input line loss feature parameter set during actual application.

[0056] As above Figure 3 and Figure 4As shown, constructing and training the line loss determination model actually uses the input sample line loss feature parameter set to train the deep belief network (DBN) model (i.e., the initial line loss determination model in the embodiment of the present invention), so that the deep belief network model (DBN) can learn the characteristics of the sample line loss feature parameter set, and then continuously train and optimize to obtain the final line loss determination model.

[0057] The Deep Belief Network (DBN) model is a network composed of multiple RBMs and a logistic regression layer, which has strong self-learning ability. The top layer is used as the output layer for classification or regression. It achieves the optimal model training through greedy forward learning and reverse fine-tuning of gradient descent, as shown above. Figure 4 As shown in the figure, the first layer of neurons and the second layer of neurons form the first RBM, the second layer of neurons and the third layer of neurons form the second RBM, and so on until the nth layer of neurons, so the DBN model has a total of n RBMs and one output layer.

[0058] Specifically, the deep belief network (DBN) model is first initialized: including the selection of each RBM (restricted Boltzmann machine) and output layer activation function, the selection of the loss function, and the determination of hyperparameters such as the number of neurons, the number of neurons in each layer, the learning rate, the momentum, and the training batch; then the deep belief network (DBN) model is pre-trained: using the sample line loss feature parameter set as input data, training is performed layer by layer from the bottom layer to the top layer of the DBN structure, and the neuron state of each layer is updated; then the pre-trained deep belief network (DBN) model is reversely fine-tuned: multiple groups of labeled sample line loss feature parameter sets (i.e., the labeled sample line loss feature parameter sets in the embodiment of the present invention) are input into the pre-trained DBN, the theoretical line loss value is used as the output, the root mean square loss function is adopted, and the Adam optimization algorithm is used to obtain the optimization parameters. When the preset accuracy is met or the preset number of iterations is reached, the training is stopped, the trained DBN model parameters are saved, and the current deep belief network (DBN) model is used as the trained line loss determination model.

[0059] In a specific embodiment of the present invention, the initial line loss determination model is iteratively trained using multiple sets of sample line loss feature parameter sets until a convergence condition is reached, and the current initial line loss determination model is determined to be a line loss determination model, including: using multiple sets of sample line loss feature parameter sets to pre-train the initial line loss determination model, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters to obtain the RBM weight parameters of each neural layer, wherein the RBM weight parameters are the weight parameters of the restricted Boltzmann machine in the neural layer; using multiple sets of labeled sample line loss feature parameter sets to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until a convergence condition is reached, and the current initial line loss determination model is determined to be a line loss determination model, wherein the labeled sample line loss feature parameter set is a sample line loss feature parameter set with a corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0060] Specifically, using the sample line loss feature parameter set to pre-train the DBN model is an unsupervised training process, that is, a process of learning from the bottom neural layer to the top neural layer. It uses the unsupervised greedy training mechanism of RBM to train layer by layer to obtain the weight parameters of each RBM, so as to train the low-level data features into high-level features layer by layer, and finally send them to the top-level output layer (regression layer) to prepare for achieving the global optimum. The purpose of pre-training is to provide a better initial value for the weight parameter to avoid the problem of gradient vanishing in the deep neural network of the gradient descent algorithm. Since unsupervised training cannot make the parameters optimal, it is necessary to use the labeled sample line loss feature parameter set to fine-tune the direction of the pre-trained DBN model through supervised training, calculate the error between the output layer (that is, the first predicted theoretical line loss calculated by the initial test line loss determination model according to the sample line loss feature parameter set) and the true label (that is, the sample theoretical line loss corresponding to the sample line loss feature parameter set), and use the gradient descent algorithm to adjust the model coefficients layer by layer to reduce the error. In other words, pre-training locates the approximate range of the optimal solution of the model, while reverse fine-tuning accurately finds the optimal solution.

[0061] In the above embodiment of the present invention, multiple sets of labeled sample line loss feature parameter sets are used to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, including: a first acquisition step, inputting the labeled sample line loss feature parameter set into the pre-trained initial line loss determination model for processing, and obtaining a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is the initial line loss determination model pre-training the labeled sample line loss feature parameter set. The predicted theoretical line loss is measured; the second acquisition step is to obtain a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; the third acquisition step is to compare the first deviation value with the deviation threshold to obtain a first comparison result; the optimization step is to optimize the reverse fine-tuning process of the initial line loss determination model using an adaptive motion estimation algorithm when the first comparison result indicates that the first deviation value is greater than the deviation threshold; the first acquisition step, the second acquisition step, the third acquisition step and the optimization step are repeated until the first deviation value is not greater than the deviation threshold, and the current initial line loss determination model is determined to be the line loss determination model.

[0062] Specifically, the stochastic gradient descent algorithm (SGD) is usually used in the reverse fine-tuning of the traditional DBN model. Since its samples are random, the loss function obtained does not necessarily lead to the global optimum. Therefore, in this embodiment, the adaptive motion estimation algorithm (Adam) is selected for reverse fine-tuning optimization. The Adam algorithm is a gradient descent algorithm that adaptively adjusts the learning rate. By retaining the exponential decay average of the previous gradient and the square gradient, it reduces the swing and speeds up the convergence speed; the Adam algorithm uses the differential square weighted average for the gradient of the weight. Assuming that in the t-1 iteration, the update formula is as follows: t =μ1m t-1 +(1-μ1)dw、n t =μ2n t-1 +(1-μ2)dw 2 , In the formula, m t is the first-order moment estimate, n t is the second-order moment estimate, μ1 is a constant generally taken as 0.9, μ2 is a constant generally taken as 0.999, To correct the first-order matrix deviation, To correct the second-order matrix deviation, w tis the corrected weight, α represents the correction constant, and ε represents the error in the training process; the Adam algorithm is used for reverse fine-tuning optimization to minimize the error between the output layer (that is, the first predicted theoretical line loss calculated by the initial line loss determination model according to the sample line loss feature parameter set) and the true label (that is, the sample theoretical line loss corresponding to the sample line loss feature parameter set) to the maximum extent; through continuous iteration until the accuracy requirement is met (that is, the deviation between the first predicted theoretical line loss and the sample theoretical line loss is less than the deviation threshold), the initial line loss determination model obtained by the current training is determined as the line loss determination model.

[0063] In another specific embodiment of the present invention, a plurality of sets of labeled sample line loss feature parameter sets are used to reverse fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, including: in the process of reverse fine-tuning the pre-trained initial line loss determination model using a plurality of sets of labeled sample line loss feature parameter sets, counting the number of iterations of the labeled sample line loss feature parameter sets; when the number of iterations reaches a preset number of iterations, determining the current initial line loss determination model to be the line loss determination model.

[0064] Specifically, in this embodiment, if the DBN model fails to meet the accuracy requirements after multiple iterations of training, in order to avoid too long training time, a maximum number of iterations (preset number of iterations) can be set in advance. If the number of iterations reaches the maximum number of iterations during the DBN model training process, the training can also be stopped, and the initial line loss determination model obtained by the current training can be determined as the line loss determination model.

[0065] According to the above embodiment of the present invention, after the initial line loss determination model is iteratively trained using multiple groups of sample line loss feature parameter sets until the convergence condition is reached and the current initial line loss determination model is determined to be the line loss determination model, the distribution network theoretical line loss determination method based on the deep belief network also includes: obtaining a test line loss feature parameter set and a test theoretical line loss corresponding to the test line loss feature parameter set; inputting the test line loss feature parameter set into the line loss determination model for processing to obtain a second predicted theoretical line loss corresponding to the test line loss feature parameter set, wherein the second predicted theoretical line loss is the predicted theoretical line loss obtained by the line loss determination model according to the test line loss feature parameter set; comparing the second deviation value between the second predicted theoretical line loss and the test theoretical line loss with the deviation threshold to obtain a second comparison result; when the second comparison result indicates that the second deviation value is less than the deviation threshold, it is determined that the line loss determination model has successfully passed the accuracy verification.

[0066] As above Figure 3As shown, the test data (including the test line loss feature parameter set and the test theoretical line loss corresponding to the test line loss feature parameter set) can be input into the trained line loss determination model to determine whether the deviation between the second predicted theoretical line loss output by the line loss determination model and the test theoretical line loss is less than the deviation threshold. If this condition is met, it is considered that the line loss determination model has passed the accuracy verification and can be put into practical application. Of course, in the actual application process of the line loss determination model, the line loss determination model can also be continuously optimized to improve the accuracy and efficiency of the line loss determination model.

[0067] It can be seen from the above that through the technical solution provided by the above embodiment of the present invention, a line loss characteristic parameter set of a distribution network can be obtained, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; multiple line loss characteristic parameters in the line loss characteristic parameter set are standardized; the standardized line loss characteristic parameter set is input into the line loss determination model, so as to process the line loss characteristic parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is a deep belief network using multiple sets of training data. The data are obtained by training in a manner, and each of the multiple groups of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set, so as to select multiple characteristic parameters of the distribution network as input variables of the line loss determination model, and then use the line loss determination model trained based on the deep belief network to process the input variables to output the theoretical line loss value. The line loss determination model constructed by the deep belief network is used to analyze the correlation between the distribution network line loss feature value and the theoretical line loss, so as to quickly and efficiently calculate the theoretical line loss value. The technical effect improves the calculation speed and accuracy of the theoretical line loss.

[0068] Therefore, the technical solution provided by the above-mentioned embodiment of the present invention solves the technical problem in the related art that it is difficult to quickly and efficiently calculate the theoretical line loss accurately when using the traditional method to calculate the theoretical line loss of the power grid.

[0069] Combine the following Figure 5 The superiority of the line loss determination model trained in the above embodiment of the present invention is verified. Figure 5 3 is a schematic diagram comparing the line loss calculation results of DBN and ANN according to an embodiment of the present invention.

[0070] First, the two algorithms (SGD and Adam) used in the reverse fine-tuning process are compared in combination with Table 1. Table 1 shows the time comparison of the two reverse fine-tuning algorithms at a certain accuracy; the influence of Adam and SGD algorithms on the training results is analyzed from the convergence accuracy and convergence speed in the distribution network line loss calculation; in terms of convergence speed, the time consumed by the two reverse fine-tuning optimization algorithms to reach 0.01 and 0.001 training target errors is used as the standard; both methods can converge to 0.01, and Adam takes less time; when the error is 0.001, SGD cannot converge, while Adam only takes 7.35 seconds. The use of Adam optimization algorithm can greatly reduce the time required for DBN line loss calculation training, so this paper chooses Adam optimization algorithm for reverse fine-tuning of DBN.

[0071] Table 1

[0072] Accuracy target SGD Adam 0.01 12.27s 5.88s 0.001 / 7.35s

[0073] Secondly, in order to verify the superiority of the line loss determination model obtained by two-stage training, the artificial neural network (ANN) calculation results are selected for comparison. The structure and parameter settings of ANN are consistent with those of DBN, except that there is no unsupervised pre-training. From Table 2 and Figure 5 It can be observed that the DBN line loss prediction model is more accurate than the ANN model. Table 2 shows the calculation accuracy comparison between DBN and ANN.

[0074] Table 2

[0075]

[0076] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0077] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0078] According to an embodiment of the present invention, there is also provided a distribution network theoretical line loss determination device based on a deep belief network for implementing the above distribution network theoretical line loss determination method based on a deep belief network. Figure 6 is a schematic diagram of a device for determining theoretical line loss of a distribution network based on a deep belief network according to an embodiment of the present invention, such as Figure 6 As shown, the device includes: a first acquisition unit 61, a first processing unit 63 and a second acquisition unit 65. The device for determining theoretical line loss of a distribution network based on a deep belief network is described in detail below.

[0079] The first acquisition unit 61 is used to acquire a line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network.

[0080] The first processing unit 63 is used to perform standardization processing on multiple line loss characteristic parameters in the line loss characteristic parameter set.

[0081] The second acquisition unit 65 is used to input the standardized line loss feature parameter set into the line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple sets of training data through a deep belief network, and each of the multiple sets of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0082] It should be noted here that the above-mentioned first acquisition unit 61, first processing unit 63 and second acquisition unit 65 correspond to steps S202 to S206 in the above-mentioned embodiments, and the three units and the corresponding steps implement the same instances and application scenarios, but are not limited to the contents disclosed in the above-mentioned embodiments.

[0083] As can be seen from the above, in the scheme recorded in the above embodiment of the present invention, the first acquisition unit can be used to acquire the line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; then the first processing unit is used to standardize the multiple line loss characteristic parameters in the line loss characteristic parameter set; finally, the second acquisition unit is used to input the standardized line loss characteristic parameter set into the line loss determination model, so as to process the line loss characteristic parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is a model that uses multiple training sets to determine the line loss characteristic parameters. The training data is obtained by training in a deep belief network manner. Each of the multiple groups of training data includes: a sample line loss feature parameter set and a sample theoretical line loss corresponding to the sample line loss feature parameter set, so as to select multiple characteristic parameters of the distribution network as input variables of the line loss determination model, and then use the line loss determination model trained based on the deep belief network to process the input variables to output the theoretical line loss value. The line loss determination model constructed by the deep belief network analyzes the correlation between the distribution network line loss feature value and the theoretical line loss, so as to quickly and efficiently calculate the theoretical line loss value. The technical effect improves the calculation speed and accuracy of the theoretical line loss.

[0084] Therefore, the technical solution provided by the above-mentioned embodiment of the present invention solves the technical problem in the related art that it is difficult to quickly and efficiently calculate the theoretical line loss accurately when using the traditional method to calculate the theoretical line loss of the power grid.

[0085] Optionally, the first processing unit includes: a first calculation module, configured to calculate an average value of each line loss characteristic parameter in the line loss characteristic parameter set according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any line loss characteristic parameter set, n represents the total number of line loss characteristic parameter sets, x represents the number of the line loss characteristic parameter sets, ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; a second calculation module, used to calculate the standard deviation of each line loss characteristic parameter in the line loss characteristic parameter set according to a second formula, wherein the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; a first determination module is used to determine the standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter according to a third formula, wherein the third formula is: X ij Represents the standard line loss characteristic parameter corresponding to the j-th line loss characteristic parameter in the i-th line loss characteristic parameter set.

[0086] Optionally, the distribution network theoretical line loss determination device based on deep belief network also includes: a third acquisition unit, which is used to obtain multiple groups of sample line loss feature parameter sets before inputting the standardized line loss feature parameter set into the line loss determination model to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein each group of sample line loss feature parameter sets in the multiple groups of sample line loss feature parameter sets includes at least the following sample line loss feature parameters: sample line length, sample line model, sample distribution quantity, sample distribution model and sample distribution power; a second processing unit, which is used to A Z-score normalization method is used to standardize multiple sample line loss feature parameters in multiple groups of sample line loss feature parameter sets; a construction unit is used to construct an initial line loss determination model based on a deep belief network; a first determination unit is used to iteratively train the initial line loss determination model using multiple groups of sample line loss feature parameter sets until a convergence condition is reached, and the current initial line loss determination model is determined to be a line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches a preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than a deviation threshold.

[0087] Optionally, the determination unit includes: an acquisition module, which is used to pre-train the initial line loss determination model using multiple groups of sample line loss feature parameter sets, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters, and the RBM weight parameters of each neural layer are obtained, wherein the RBM weight parameters are the weight parameters of the restricted Boltzmann machine in the neural layer; a second determination module, which is used to reversely fine-tune the pre-trained initial line loss determination model using multiple groups of labeled sample line loss feature parameter sets, so as to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, wherein the labeled sample line loss feature parameter set is a sample line loss feature parameter set with a corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0088] Optionally, the second determination module includes: a first acquisition submodule, which is used to input the labeled sample line loss feature parameter set into the pre-trained initial line loss determination model for processing, and obtain a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is the predicted theoretical line loss obtained by the initial line loss determination model predicting the labeled sample line loss feature parameter set; a second acquisition submodule, which is used to obtain a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; a third acquisition submodule, which is used to compare the first deviation value with the deviation threshold to obtain a first comparison result; an optimization submodule, which is used to optimize the reverse fine-tuning process of the initial line loss determination model using an adaptive motion estimation algorithm when the first comparison result indicates that the first deviation value is greater than the deviation threshold; the first determination submodule, which is used to repeatedly execute the first acquisition submodule, the second acquisition submodule, the third acquisition submodule and the optimization submodule until the first deviation value is not greater than the deviation threshold, and the current initial line loss determination model is determined to be the line loss determination model.

[0089] Optionally, the second determination module includes: a statistical submodule, used to count the number of iterations of the labeled sample line loss feature parameter set in the process of reverse fine-tuning the pre-trained initial line loss determination model using multiple sets of labeled sample line loss feature parameter sets; and a second determination submodule, used to determine that the current initial line loss determination model is the line loss determination model when the number of iterations reaches a preset number of iterations.

[0090] Optionally, the distribution network theoretical line loss determination device based on a deep belief network also includes: a fourth acquisition unit, which is used to iteratively train the initial line loss determination model using multiple groups of sample line loss feature parameter sets until the convergence condition is reached, and then determine that the current initial line loss determination model is the line loss determination model, and then obtain a test line loss feature parameter set and a test theoretical line loss corresponding to the test line loss feature parameter set; a fifth acquisition unit, which is used to input the test line loss feature parameter set into the line loss determination model for processing, and obtain a second predicted theoretical line loss corresponding to the test line loss feature parameter set, wherein the second predicted theoretical line loss is the predicted theoretical line loss obtained by the line loss determination model according to the test line loss feature parameter set; a sixth acquisition unit, which is used to compare the second deviation value between the second predicted theoretical line loss and the test theoretical line loss with a deviation threshold to obtain a second comparison result; and a second determination unit, which is used to determine that the line loss determination model has successfully passed the accuracy verification when the second comparison result indicates that the second deviation value is less than the deviation threshold.

[0091] According to another aspect of an embodiment of the present invention, a distribution network theoretical line loss determination system based on a deep belief network is also provided. The distribution network theoretical line loss determination system based on a deep belief network uses any of the above-mentioned distribution network theoretical line loss determination methods based on a deep belief network.

[0092] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided, the computer-readable storage medium comprising a stored program, wherein the program executes any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network.

[0093] Optionally, in this embodiment, the computer-readable storage medium may be located in any one of the computer terminals in a computer terminal group in a computer network, or in any one of the communication devices in a communication device group.

[0094] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: obtaining a line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; standardizing the multiple line loss characteristic parameters in the line loss characteristic parameter set; inputting the standardized line loss characteristic parameter set into a line loss determination model, so as to process the line loss characteristic parameters using the line loss determination model to obtain a theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple sets of training data by means of a deep belief network, and each of the multiple sets of training data includes: a sample line loss characteristic parameter set, and a sample theoretical line loss corresponding to the sample line loss characteristic parameter set.

[0095] Optionally, in this embodiment, the computer-readable storage medium is configured to store a program code for executing the following steps: calculating the average value of each line loss characteristic parameter in the line loss characteristic parameter set according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any line loss characteristic parameter set, n represents the total number of line loss characteristic parameter sets, x represents the number of the line loss characteristic parameter sets, ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; the standard deviation of each line loss characteristic parameter in the line loss characteristic parameter set is calculated according to the second formula, where the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; and determines the standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter according to the third formula, wherein the third formula is: X ijRepresents the standard line loss characteristic parameter corresponding to the j-th line loss characteristic parameter in the i-th line loss characteristic parameter set.

[0096] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for executing the following steps: obtaining multiple groups of sample line loss feature parameter sets, wherein each group of sample line loss feature parameter sets in the multiple groups of sample line loss feature parameter sets includes at least the following sample line loss feature parameters: sample line length, sample line model, sample distribution quantity, sample distribution model and sample distribution quantity; using Z-score normalization method to standardize multiple sample line loss feature parameters in the multiple groups of sample line loss feature parameter sets; constructing an initial line loss determination model based on a deep belief network; using multiple groups of sample line loss feature parameter sets to iteratively train the initial line loss determination model until a convergence condition is reached, determining the current initial line loss determination model as the line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches a preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than a deviation threshold.

[0097] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for performing the following steps: using multiple groups of sample line loss feature parameter sets to pre-train the initial line loss determination model, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters to obtain the RBM weight parameters of each neural layer, wherein the RBM weight parameters are the weight parameters of the restricted Boltzmann machine in the neural layer; using multiple groups of labeled sample line loss feature parameter sets to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, wherein the labeled sample line loss feature parameter set is a sample line loss feature parameter set with a corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

[0098] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for executing the following steps: a first acquisition step, inputting the labeled sample line loss feature parameter set into the pre-trained initial line loss determination model for processing, and obtaining a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is the predicted theoretical line loss obtained by the initial line loss determination model predicting the labeled sample line loss feature parameter set; a second acquisition step, obtaining a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; a third acquisition step, comparing the first deviation value with a deviation threshold to obtain a first comparison result; an optimization step, when the first comparison result indicates that the first deviation value is greater than the deviation threshold, optimizing the reverse fine-tuning process of the initial line loss determination model using an adaptive motion estimation algorithm; repeating the first acquisition step, the second acquisition step, the third acquisition step and the optimization step until the first deviation value is not greater than the deviation threshold, determining that the current initial line loss determination model is the line loss determination model.

[0099] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for performing the following steps: in the process of reverse fine-tuning the pre-trained initial line loss determination model using multiple sets of labeled sample line loss feature parameter sets, counting the number of iterations of the labeled sample line loss feature parameter sets; when the number of iterations reaches a preset number of iterations, determining that the current initial line loss determination model is the line loss determination model.

[0100] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for executing the following steps: obtaining a test line loss characteristic parameter set and a test theoretical line loss corresponding to the test line loss characteristic parameter set; inputting the test line loss characteristic parameter set into the line loss determination model for processing to obtain a second predicted theoretical line loss corresponding to the test line loss characteristic parameter set, wherein the second predicted theoretical line loss is the predicted theoretical line loss obtained by the line loss determination model based on the test line loss characteristic parameter set; comparing a second deviation value between the second predicted theoretical line loss and the test theoretical line loss with a deviation threshold to obtain a second comparison result; when the second comparison result indicates that the second deviation value is less than the deviation threshold, determining that the line loss determination model has successfully passed the accuracy verification.

[0101] According to another aspect of an embodiment of the present invention, a processor is further provided, and the processor is used to run a program, wherein when the program is running, any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network is executed.

[0102] According to another aspect of an embodiment of the present invention, a computer program product is also provided, comprising computer instructions, which, when executed by a processor, execute any one of the above-mentioned methods for determining theoretical line loss of a distribution network based on a deep belief network.

[0103] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0104] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0106] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0107] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0108] If the integrated unit 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 invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of 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 invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0109] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for determining theoretical line loss in distribution networks based on deep belief networks. It is characterized in that include: Acquire a line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; Performing standardization processing on a plurality of the line loss characteristic parameters in the line loss characteristic parameter set; The standardized line loss feature parameter set is input into the line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained using multiple groups of training data through a deep belief network, and each of the multiple groups of training data includes: a sample line loss feature parameter set, and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

2. According to the method for determining theoretical line loss of distribution network based on deep belief network according to claim 1, It is characterized in that The method of standardizing the plurality of line loss characteristic parameters in the line loss characteristic parameter set comprises: The average value of each of the line loss characteristic parameters in the line loss characteristic parameter set is calculated according to a first formula, wherein the first formula is: i represents the number of the line loss characteristic parameter set, j represents the number of the line loss characteristic parameter in any of the line loss characteristic parameter sets, n represents the total number of the line loss characteristic parameter sets, x ij represents the jth line loss characteristic parameter in the i-th line loss characteristic parameter set, represents the average value of the j-th line loss characteristic parameter; The standard deviation of each of the line loss characteristic parameters in the line loss characteristic parameter set is calculated according to a second formula, wherein the second formula is: s ij represents the standard deviation of the jth line loss characteristic parameter in the i-th line loss characteristic parameter set; The standard line loss characteristic parameter obtained by standardizing the line loss characteristic parameter is determined according to a third formula, wherein the third formula is: X ij Represents the standard line loss characteristic parameter corresponding to the jth line loss characteristic parameter in the i-th line loss characteristic parameter set.

3. According to the method for determining theoretical line loss of distribution network based on deep belief network according to claim 1, It is characterized in that Before inputting the standardized line loss characteristic parameter set into the line loss determination model to process the line loss characteristic parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, the method further includes: Acquire multiple groups of sample line loss feature parameter sets, wherein each group of sample line loss feature parameter sets in the multiple groups of sample line loss feature parameter sets includes at least the following sample line loss feature parameters: sample line length, sample line model, sample power distribution quantity, sample power distribution model and sample power distribution quantity; Using a Z-score standardization method to standardize a plurality of the sample line loss characteristic parameters in a plurality of groups of the sample line loss characteristic parameter sets; Constructing an initial line loss determination model based on the deep belief network; The initial line loss determination model is iteratively trained using multiple sets of sample line loss feature parameter sets until a convergence condition is reached, and the current initial line loss determination model is determined to be the line loss determination model, wherein the convergence condition is at least one of the following: the number of iterations reaches a preset number of iterations, and the deviation between the sample theoretical line loss output by the initial line loss determination model and the corresponding actual theoretical line loss is not greater than a deviation threshold.

4. According to the method for determining theoretical line loss of distribution network based on deep belief network according to claim 3, It is characterized in that Iteratively training the initial line loss determination model using multiple sets of the sample line loss feature parameter sets until a convergence condition is reached, and determining the current initial line loss determination model as the line loss determination model, including: Pre-training the initial line loss determination model using multiple sets of the sample line loss feature parameter sets, so that the initial line loss determination model is trained layer by layer from the bottom neural layer to the top neural layer based on the sample line loss feature parameters, and an RBM weight parameter of each neural layer is obtained, wherein the RBM weight parameter is a weight parameter of a restricted Boltzmann machine in the neural layer; Use multiple groups of labeled sample line loss feature parameter sets to reversely fine-tune the initial line loss determination model after pre-training to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, and determine that the current initial line loss determination model is the line loss determination model, wherein the labeled sample line loss feature parameter set is the sample line loss feature parameter set with the corresponding sample theoretical line loss label, and the sample theoretical line loss label is used to represent the sample theoretical line loss corresponding to the sample line loss feature parameter set.

5. According to the method for determining theoretical line loss of distribution network based on deep belief network according to claim 4, It is characterized in that Using a plurality of sets of labeled sample line loss feature parameter sets to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, determining the current initial line loss determination model as the line loss determination model, including: A first acquisition step is to input the labeled sample line loss feature parameter set into the pre-trained initial line loss determination model for processing, so as to obtain a first predicted theoretical line loss corresponding to the labeled sample line loss feature parameter set, wherein the first predicted theoretical line loss is the predicted theoretical line loss obtained by the initial line loss determination model predicting the labeled sample line loss feature parameter set; A second acquisition step of acquiring a first deviation value between the first predicted theoretical line loss and the sample theoretical line loss corresponding to the labeled sample line loss feature parameter set; A third acquisition step is to compare the first deviation value with the deviation threshold to obtain a first comparison result; an optimization step, when the first comparison result indicates that the first deviation value is greater than the deviation threshold, optimizing the reverse fine-tuning process of the initial line loss determination model by using an adaptive motion estimation algorithm; The first acquisition step, the second acquisition step, the third acquisition step and the optimization step are repeatedly performed until the first deviation value is not greater than the deviation threshold, and the current initial line loss determination model is determined to be the line loss determination model.

6. According to the method for determining theoretical line loss of distribution network based on deep belief network according to claim 4, It is characterized in that Using a plurality of sets of labeled sample line loss feature parameter sets to reversely fine-tune the pre-trained initial line loss determination model to adjust the RBM weight parameters of each neural layer in the initial line loss determination model until the convergence condition is reached, determining the current initial line loss determination model as the line loss determination model, including: In the process of reversely fine-tuning the pre-trained initial line loss determination model using the plurality of sets of the labeled sample line loss feature parameter sets, counting the number of iterations of the labeled sample line loss feature parameter sets; When the number of iterations reaches a preset number of iterations, the current initial line loss determination model is determined as the line loss determination model.

7. The method for determining theoretical line loss of a distribution network based on a deep belief network according to claim 3, It is characterized in that After iteratively training the initial line loss determination model using the plurality of sets of sample line loss feature parameter sets until a convergence condition is reached and determining that the current initial line loss determination model is the line loss determination model, the method further includes: Acquire a test line loss characteristic parameter set and a test theoretical line loss corresponding to the test line loss characteristic parameter set; Inputting the test line loss characteristic parameter set into the line loss determination model for processing to obtain a second predicted theoretical line loss corresponding to the test line loss characteristic parameter set, wherein the second predicted theoretical line loss is a predicted theoretical line loss obtained by the line loss determination model according to the test line loss characteristic parameter set; Compare a second deviation value between the second predicted theoretical line loss and the tested theoretical line loss with the deviation threshold to obtain a second comparison result; When the second comparison result indicates that the second deviation value is less than the deviation threshold, it is determined that the line loss determination model has successfully passed the accuracy verification.

8. A device for determining theoretical line loss of distribution network based on deep belief network, It is characterized in that include: A first acquisition unit is used to acquire a line loss characteristic parameter set of the distribution network, wherein the line loss characteristic parameter set is a collection of multiple line loss characteristic parameters, and each of the multiple line loss characteristic parameters is used to represent the loss characteristics of the distribution network; A first processing unit, configured to perform standardization processing on a plurality of the line loss characteristic parameters in the line loss characteristic parameter set; The second acquisition unit is used to input the line loss feature parameter set after standardization into the line loss determination model, so as to process the line loss feature parameters using the line loss determination model to obtain the theoretical line loss of the distribution network, wherein the line loss determination model is trained by using multiple groups of training data through a deep belief network, and each of the multiple groups of training data includes: a sample line loss feature parameter set and a sample theoretical line loss corresponding to the sample line loss feature parameter set.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium includes a stored program, wherein the program executes the method for determining theoretical line loss of a distribution network based on a deep belief network as described in any one of claims 1 to 7.

10. A computer program product comprising computer instructions, It is characterized in that When the computer instructions are executed by the processor, the method for determining theoretical line loss of a distribution network based on a deep belief network as described in any one of claims 1 to 7 is executed.