Middle and short term line loss rate prediction method and system based on VMD-EWT-LSTM, electronic equipment and medium
The VMD-EWT-LSTM method predicts the power supply and power sales, which solves the problem of large errors in line loss rate prediction, achieves higher-precision line loss rate prediction, and supports grid optimization management.
Patent Information
- Application Number
- CN202510378328.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-28
- Publication Date
- 2025-07-25
AI Technical Summary
The existing line loss rate prediction methods will lead to large percentage errors under slight absolute errors, and the prediction accuracy of a single machine learning model is limited, making it difficult to support the formulation of loss reduction solutions in the future stage.
The VMD-EWT-LSTM-based method is used to reduce the time series complexity through VMD decomposition and EWT decomposition, and predict the power supply and power sales volume with the LSTM model, and indirectly calculate the future line loss rate.
It improves the accuracy of line loss rate prediction, simplifies the model construction process, can more accurately predict line loss rates in the future stage, and supports more effective formulation of loss reduction solutions.
Smart Images
Figure CN120373526A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power grids, and in particular to a medium- and short-term line loss rate prediction method, system, electronic device and medium based on VMD-EWT-LSTM. Background Art
[0002] Grid line loss is an important comprehensive technical and economic indicator of power grid enterprises, and has become one of the key comprehensive indicators for measuring the operation quality, transmission efficiency, management level, etc. of the power grid, and is also the theoretical basis for formulating power grid construction, operation dispatching, control decisions, etc. In order to fully tap the potential of regional loss reduction and more effectively formulate loss reduction plans, a large number of scholars have carried out research on line loss rate prediction, and the main methods include statistical models (such as k-nearest neighbor, Kalman filter, etc.) and machine learning models (support vector product, recurrent neural network, long short-term memory network).
[0003] At present, the line loss rate prediction method usually establishes feature engineering such as meteorology and date, and combines historical line loss data with machine learning algorithms to realize the prediction of future line loss rate. However, the line loss rate is usually a relatively small value, and a small absolute error will cause a large percentage error. Therefore, directly predicting the line loss rate based on historical line loss data will result in an unsatisfactory prediction effect of the model; in addition, the prediction accuracy of a single machine learning model is limited, and its prediction results are difficult to support the formulation of loss reduction plans in the future stage. Summary of the Invention
[0004] The purpose of the embodiments of the present invention is to provide a medium- and short-term line loss rate prediction method, system, electronic device and medium based on VMD-EWT-LSTM, which can greatly reduce the complexity of the original time series and has higher prediction accuracy compared with general prediction models.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] In a first aspect, the embodiments of the present invention provide a medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM, including the following steps:
[0007] Collect historical power supply data and historical power sales data of the target area, and represent them with GD ij , SD ij respectively;
[0008] Construct a power supply amount prediction model for the target area at a future moment based on the collected historical power supply data of the target area;
[0009] Construct a power sales amount prediction model for the target area at a future moment based on the collected historical power sales data of the target area;
[0010] Predict the power supply of the target area at a future moment based on the power supply prediction model of the target area at a future moment, predict the future moment of the target area based on the power sales prediction model of the target area at a future moment, and calculate the line loss rate of the target area in the future stage based on the predicted power supply and power sales of the target area at a future moment.
[0011] Further, collect the historical power supply data and historical power sales data of the target area, and use GD ij and SD ij to represent specifically as follows:
[0012] i represents the specific date of the data, and j represents the specific moment of the data. Set the data collection range to I days and 24 o'clock data to obtain GD ij and SD ij The subscript value ranges are as follows:
[0013] i ∈ {1, 2, 3,......, I}(1)
[0014] j ∈ {1, 2, 3,......, 24}(2).
[0015] Further, the specific method for constructing the power supply prediction model of the target area at a future moment based on the collected historical power supply data of the target area is as follows:
[0016] (1) VMD decomposition
[0017] ① Regard the historical power supply data GD ij of the target area as a continuous time signal. Assume that it can be decomposed by VMD into k primary subsequences. Then the expression of the kth primary subsequence is:
[0018] u k (t) = A k (t)cos[φ k (t)], k ∈ {1, 2,..., K}(3)
[0019] In the above formula, the value range of t is determined by the product of the numbers of i and j, that is, the number of primary subsequences is consistent with the number of historical power supply data of the target area. Therefore, GD ij is replaced by GD t ; the phase φ k (t) is a non-decreasing function, and φ k '(t) ≥ 0; A k (t) represents the envelope function;
[0020] ② The bandwidth of each IMF component can be estimated according to Carson's criterion, specifically as follows:
[0021] BW AM-FM = 2(Δf + f FM+f AM )(4)
[0022] In the above formula, Δf represents the maximum deviation of the instantaneous frequency from the center, and f FM represents the highest frequency of the envelope function A k (t);
[0023] ③ Under the constraint condition that the sum of each component is equal to the input signal, minimize the sum of the estimated bandwidths of each component, and then through a series of transformations, construct the following constrained variational model:
[0024]
[0025] In the formula: {u k (t)} = {u1(t), u2(t),......, u k (t),......, u K (t)} is the set of each primary subsequence; {w k} = {w1, w2,......, w k ,......, w K} is the set of the corresponding center frequencies; is the unit impulse function;
[0026] ④ According to the historical power supply data GD t of the original target area and the set of the primary subsequence {u k (t)}, solve the residual sequence re(t):
[0027]
[0028] (2) EWT decomposition
[0029] ① Transform the residual sequence re(t) into the sequence R(w) through FFT, and then divide the spectrum of the sequence R(w) into N continuous segments; where, FFT refers to the fast Fourier transform; the sequence R(w) is the representation after the residual sequence re(t) undergoes the fast Fourier transform;
[0030] ② Define the empirical scaling function φ s (w) and the empirical wavelet function ψ s (w) in the band-pass filter of EWT:
[0031]
[0032]
[0033] Among them, w represents the frequency,
[0034]
[0035] z in Equation (9) is a random number;
[0036] ③ Calculate the inner product of the residual sequence re(t) and the empirical scaling function φ s (w), and the inner product of the residual sequence re(t) and the empirical wavelet function ψ s (w), respectively obtaining the approximation coefficient W f (0,t) and the detail coefficient W f (s,t):
[0037]
[0038] ④ Calculate the quadratic subsequence e m (t)
[0039]
[0040] where denotes the convolution calculation, m = 1, 2, 3,..., M, e1(t) represents the first quadratic subsequence, and e m (t) represents the m-th quadratic subsequence;
[0041] (3) LSTM Short-Term Power Supply Data Prediction
[0042] ① For the first primary subsequence u1(t), use LSTM to conduct short-term prediction, and the prediction result is denoted as u1'(t).
[0043]
[0044] In Equation (13), f t , i t , c t represent the forget gate, input gate, and output gate respectively, h t represents the t-th output, h t-1 represents the (t - 1)-th output, c t represents the cell state, represents the candidate cell state, w and b represent the weight vector and bias vector, and σ and tanh represent the activation functions;
[0045] The constructed LSTM neural network has a total of five layers. The first layer is the first LSTM layer, with 200 neurons set; the second layer is the Dropout layer, with the parameter set to 0.4; the third layer is the second LSTM layer, with 240 neurons set; the fourth layer is the Dropout layer, with the parameter set to 0.4; the last layer is the Dense layer, with the parameter set to 1;
[0046] Using a for loop, the first to 60th data of each subsequence are used as the first group of inputs, and the 61st data is used as the label of the first group of data; the 2nd to 61st data are used as the second group of inputs, and the 62nd data is used as the label of the second group of data; and so on, until the (T - 60)th to (T - 1)th data are used as inputs, and the Tth input is used as the label of the last group of data. The loop ends, and the number of iterations of the model is set to 200;
[0047] Through the above parameter settings, the training of the LSTM model is realized. When inputting the subsequence u1(t) once, the prediction result u1'(t) of the subsequence u1(t) is automatically obtained;
[0048] ② Referring to the above step ①, the prediction of the future short-term numerical values of all primary subsequences and all secondary subsequences is realized;
[0049] ③ Summing up the prediction results of all primary subsequences and secondary subsequences, the prediction of the power supply quantity in the short-term target area of the "future state" is obtained, as shown in the following formula:
[0050]
[0051] In the above formula (14), u k '(t) represents the prediction result of the kth primary subsequence, e m '(t) represents the prediction result of the mth secondary subsequence, and GD t ' represents the prediction result of the power supply quantity in the short-term target area of the "future state".
[0052] Furthermore, the specific method for constructing the future moment target area electricity sales prediction model based on the collected historical electricity sales data of the target area is that the construction steps are similar to those of the power supply quantity prediction model, but the difference lies in the output of the model. The final output, the prediction result of the short-term target area electricity sales in the "future state", is represented by SD t '.
[0053] Furthermore, the calculation of the future stage target area line loss rate based on the predicted future moment target area power supply quantity and electricity sales is specifically as follows:
[0054] Based on the prediction results of the future moment target area power supply quantity and electricity sales, the calculation of the future stage target area line loss rate is realized, and it is represented by FXS t .,
[0055]
[0056] In the above formula, GD t ' represents the prediction result of the short-term target area power supply quantity in the "future state"; SD t ' represents the prediction result of the short-term target area electricity sales in the "future state"; FXSt Indicates the predicted result of the line loss rate in the short-term target area of the "future state".
[0057] In a second aspect, an embodiment of the present invention provides a medium- and short-term line loss rate prediction system based on VMD-EWT-LSTM, including:
[0058] A data collection module for collecting historical power supply data and historical power sales data of the target area, denoted by GD ij and SD ij respectively.
[0059] A power supply prediction model module for constructing a power supply prediction model of the target area at a future moment based on the collected historical power supply data of the target area;
[0060] A power sales prediction model module for constructing a power sales prediction model of the target area at a future moment based on the collected historical power sales data of the target area;
[0061] A calculation module for predicting the power supply of the target area at a future moment based on the power supply prediction model of the target area at a future moment, predicting the power sales of the target area at a future moment based on the power sales prediction model of the target area at a future moment, and calculating the line loss rate of the target area in a future stage based on the predicted power supply and power sales of the target area at a future moment.
[0062] In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM as described above are implemented.
[0063] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM as described above are implemented.
[0064] Compared with the prior art, the beneficial effects of the present invention are:
[0065] 1. Abandon the construction of feature engineering, and only construct a prediction model for one-dimensional data, making the model construction simpler.
[0066] 2. Combine the time series decomposition method and LSTM to realize the prediction of power supply and power sales in the future stage, which can greatly reduce the complexity of the original time series. Compared with general prediction models, it has higher prediction accuracy.
[0067] 3. Aiming at the disadvantage that the prediction effect of the model is not ideal for directly predicting the line loss rate based on historical line loss data, the "future state" line loss rate prediction result is indirectly calculated using the predicted results of power supply and sales electricity, making up for the deficiency that the line loss rate data is small and a small absolute error will cause a large percentage error. Description of the Drawings
[0068] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.
[0069] Figure 1 is a flowchart of a medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM according to an embodiment of the present invention.
[0070] Figure 2 is a block diagram of a module of a medium- and short-term line loss rate prediction system based on VMD-EWT-LSTM according to an embodiment of the present invention. Detailed Embodiments
[0071] The technical solutions in the embodiments of the present invention will be described below in conjunction with the drawings in the embodiments of the present invention. It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.
[0072] The term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or also elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.
[0073] The terms "first", "second", etc. are only used to distinguish one entity or operation from another entity or operation, and cannot be understood as indicating or implying relative importance, nor can it be understood as requiring or implying any actual relationship or order between these entities or operations.
[0074] As Figure 1 shown, a medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM includes the following steps:
[0075] 1. Collect the historical power supply and sales volume data of the target area, denoted by GD ij and SD ij respectively. Among them, the subscript i represents the specific date of the data, and j represents the specific moment of the data. Set the data collection range to the data of I days and 24 points, and obtain GD ij and SD ij with the subscript value ranges as follows:
[0076] i ∈ {1, 2, 3,......, I} (1)
[0077] j ∈ {1, 2, 3,......, 24} (2)
[0078] 2. Build a power supply prediction model for the target area at future moments.
[0079] (1) VMD decomposition
[0080] ①Regard the historical power supply data GD ij of the target area as a continuous time signal. Assume that it can be decomposed by VMD into k first-order subsequences. Then the expression of the k-th first-order subsequence is:
[0081] u k (t) = A k (t)cos[φ k (t)], k ∈ {1, 2,..., K} (3)
[0082] In the above formula, the value range of t is determined by the product of the numbers of i and j, that is, the number of first-order subsequences is the same as the number of historical power supply data of the target area. Therefore, GD ij can be replaced by GD t ; the phase φ k (t) is a non-decreasing function, and φ k '(t) ≥ 0; A k (t) represents the envelope function.
[0083] ②The bandwidth of each IMF component can be estimated according to Carson's criterion, as follows:
[0084] BW AM-FM = 2(Δf + f FM + f AM )(4)
[0085] In the above formula, Δf represents the maximum deviation of the instantaneous frequency from the center, and f FM represents the highest frequency of the envelope function A k (t).
[0086] ③Under the constraint condition that the sum of each component is equal to the input signal, the sum of the estimated bandwidths of each component is minimized, and after a series of transformations, the following constrained variational model is constructed:
[0087]
[0088] In the formula: {u k (t)} = {u1(t), u2(t),......, u k (t),......, u K (t)} is the set of each primary subsequence; {w k} = {w1, w2,......, w k ,......, w K} is the corresponding set of center frequencies; is the unit impulse function.
[0089] ④According to the historical power supply data GD of the original target area t and the set of primary subsequences {u k (t)}, the residual sequence re(t) is solved
[0090]
[0091] (2) EWT decomposition
[0092] ①The residual sequence re(t) is transformed into the sequence R(w) through FFT, and then the spectrum of the sequence R(w) is divided into N continuous segments; where, FFT refers to the fast Fourier transform; the sequence R(w) is the representation after the residual sequence re(t) passes through the fast Fourier transform.
[0093] ②Define the empirical scaling function φ s (w) and the empirical wavelet function ψ s (w) in the band-pass filter of EWT.
[0094]
[0095] Among them, w represents frequency,
[0096]
[0097] It should be noted that z in the above formula (9) is a random number.
[0098] ③Calculate the inner product of the residual sequence re(t) and the empirical scaling function φ s (w), and the inner product of the residual sequence re(t) and the empirical wavelet function ψ s (w), and the approximation coefficients W f(0, t) and the detail coefficient W f (s, t).
[0099]
[0100] ④ Calculate the secondary subsequence e m (t)
[0101]
[0102] In the above formula, represents the convolution calculation, m = 1, 2, 3,..., M, e1(t) represents the first secondary subsequence, and e m (t) represents the m-th secondary subsequence.
[0103] (3) LSTM short-term power supply data prediction.
[0104] ① For the first primary subsequence u1(t), use LSTM to conduct short-term prediction, and the prediction result is represented by u1'(t).
[0105]
[0106] In the above formula (13), f t 、i t 、c t represent the forget gate, input gate, and output gate respectively, h t represents the t-th output, h t-1 represents the (t - 1)-th output, c t represents the cell state, represents the candidate cell state, w and b represent the weight vector and bias vector, and σ and tanh represent the activation functions.
[0107] In this step, the constructed LSTM neural network has five layers. The first layer is the first LSTM layer, with 200 neurons set; the second layer is the Dropout layer, with the parameter set to 0.4; the third layer is the second LSTM layer, with 240 neurons set; the fourth layer is the Dropout layer, with the parameter set to 0.4; the last layer is the Dense layer. Since only one-dimensional data needs to be output, it is set to 1.
[0108] In this step, use a for loop to take the first to 60th data of each subsequence as the first group of inputs, and the 61st data as the label of the first group of data; take the 2nd to 61st data as the second group of inputs, and the 62nd data as the label of the second group of data; and so on, until the (T - 60)-th to (T - 1)-th data are used as inputs, and the T-th input is used as the label of the last group of data, and the loop ends. The number of iterations of the model is set to 200.
[0109] Through the above parameter settings, the training of the LSTM model is realized. By inputting the subsequence u1(t) once, the prediction result u1'(t) of the subsequence u1(t) can be automatically obtained.
[0110] ② Referring to step ① above, realize the prediction of the future short-term values of all primary subsequences and all secondary subsequences.
[0111] ③ Sum the prediction results of all primary subsequences and secondary subsequences to obtain the prediction of the power supply in the "future state" short-term target area, as shown in the following formula:
[0112]
[0113] In the above formula (14), u k '(t) represents the prediction result of the kth primary subsequence, and e m '(t) represents the prediction result of the mth secondary subsequence. GD t ' represents the prediction result of the power supply in the "future state" short-term target area.
[0114] 2. Build a prediction model for the electricity sales volume in the target area at future moments.
[0115] The construction steps are similar to those of the power supply prediction model, but the difference lies in the output of the model. The final output is the prediction result of the electricity sales volume in the "future state" short-term target area, denoted by SD t '.
[0116] 3. Calculate the line loss rate in the target area in the future stage.
[0117] Based on the prediction results of the power supply and electricity sales volume in the target area at future moments, calculate the line loss rate in the target area in the future stage, denoted by FXS(t).
[0118]
[0119] 4. Example demonstration
[0120] Taking the distribution network in several regions of Hubei Province as an example, combined with the line loss rate prediction method in the present invention and the real data of the power grid line loss, verify the prediction performance of the model:
[0121] Table 1 Statistics of the correct recognition rate of the distribution transformer gear positions in a certain area of Hubei Province
[0122]
[0123]
[0124] As can be seen from the content in Table 1, the future stage line loss rate prediction method in the present invention has a high accuracy rate and has certain practical application value.
[0125] As Figure 2 shown, an embodiment of the present invention provides a medium- and short-term line loss rate prediction system based on VMD-EWT-LSTM, including,
[0126] A data collection module for collecting historical power supply data and historical power sales data of a target area, represented by GD ij and SD ij respectively;
[0127] A power supply prediction model module for constructing a power supply prediction model of the target area at a future moment based on the collected historical power supply data of the target area;
[0128] A power sales prediction model module for constructing a power sales prediction model of the target area at a future moment based on the collected historical power sales data of the target area;
[0129] A calculation module for predicting the power supply of the target area at a future moment based on the power supply prediction model of the target area at a future moment, predicting the target area at a future moment based on the power sales prediction model of the target area at a future moment, and calculating the line loss rate of the target area in a future stage based on the predicted power supply and power sales of the target area at a future moment.
[0130] An embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM as described above are implemented.
[0131] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM as described above are implemented.
[0132] Those skilled in the art should understand that the embodiments of the present invention can provide a method, a system, or a computer program product. Therefore, the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0133] The present invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processors of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing device create means for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks or blocks.
[0134] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks or blocks.
[0135] These computer program instructions may also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks or blocks.
[0136] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.
[0137] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.
[0138] A computer-readable medium includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media such as modulated data signals and carrier waves.
[0139] The above are only examples of the embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.
Claims
1. A medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM, characterized in that It includes the following steps: Collect the historical power supply data and historical power sales data of the target area, denoted by GD ij and SD ij respectively; Construct a power supply quantity prediction model for the target area at a future moment based on the collected historical power supply quantity data of the target area; Construct a power sales quantity prediction model for the target area at a future moment based on the collected historical power sales quantity data of the target area; Predict the power supply quantity of the target area at a future moment based on the power supply quantity prediction model for the target area at a future moment, predict the target area at a future moment based on the power sales quantity prediction model for the target area at a future moment, and calculate the line loss rate of the target area in the future stage based on the predicted power supply quantity and power sales quantity of the target area at a future moment.
2. The short-term and medium-term line loss rate prediction method based on VMD-EWT-LSTM according to claim 1, characterized in that The historical power supply data and historical power sales data of the target area are collected, and are represented by GD ij and SD ij specifically as follows: Let \(i\) represent the specific date of the data and \(j\) represent the specific time of the data. Set the data collection range to 24 - point data for \(I\) days to obtain \(GD\) ij and \(SD\) ij The subscript value range is as follows: i ∈ {1, 2, 3,......, I} (1) j∈{1,2,3,......,24} (2)。 3. A medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM according to claim 1, characterized in that The specific method for constructing a power supply quantity prediction model for the target area at a future moment based on the collected historical power supply quantity data of the target area is as follows: (1) VMD decomposition ①Regarding the historical power supply data GD of the target area ij as a continuous time signal, assuming that it can be decomposed into k first-order subsequences by VMD, the expression of the kth first-order subsequence is as follows: u k (t) = A k (t) cos[φ k (t)], k ∈ {1, 2,..., K} (3) In the above formula, the value range of t is determined by the product of the numbers of i and j, that is, the number of one - time subsequences is consistent with the number of historical power supply data in the target area. Therefore, GD ij is replaced by GD t ; the phase φ k (t) is a non - decreasing function, and φ k '(t) ≥ 0; A k (t) represents the envelope function; ② The bandwidth of each IMF component can be estimated according to Carson's criterion, as follows: BW AM-FM = 2(Δf + f FM + f AM ) (4) In the above formula, Δf represents the maximum deviation of the instantaneous frequency from the center, and f FM represents the highest frequency of the envelope function A k (t); ③ Under the constraint condition that the sum of each component is equal to the input signal, minimize the sum of the estimated bandwidths of each component, and then through a series of transformations, construct the following constrained variational model: where: {u k (t)} = {u1(t), u2(t),......, u k (t),......, u K (t)} is the set of each first-order subsequence; {w k} = {w1, w2,......, w k ,......, w K} is the set of corresponding center frequencies; is the unit impulse function; ④Solve the residual sequence re(t) based on the historical power supply data GD of the original target area t and the set of the first-order subsequences {u k (t)}: (2) EWT decomposition ① Transform the residual sequence re(t) into a sequence R(w) through FFT, and then divide the spectrum of the sequence R(w) into N continuous segments; where, FFT refers to the fast Fourier transform; the sequence R(w) is the representation of the residual sequence re(t) after the fast Fourier transform; ② Define the empirical scaling function φ s (w) and the empirical wavelet function ψ s (w) in the band-pass filter of EWT: s (w) and the empirical wavelet function ψ s (w): where \(w\) represents frequency, \(s\in[1,N]\); \(\tau\) s =\(\gamma w\) s ; z in formula (9) is a random number; ③ Calculate the inner product of the residual sequence re(t) and the empirical scaling function φ s (w), and the inner product of the residual sequence re(t) and the empirical wavelet function ψ s (w), respectively obtaining the approximation coefficient W f (0, t) and the detail coefficient W f (s, t): ④ Calculate the secondary subsequence e m (t) In the formula, represents a convolution calculation, m = 1, 2, 3, ..., M, e1(t) represents the first secondary subsequence, and e m (t) represents the m-th secondary subsequence; (3) LSTM short-term power supply data prediction ① For the first primary subsequence u1(t), use LSTM to carry out short-term prediction, and the prediction result is represented by u1'(t). f t = σ(w f * [h t-1 , u t + b f ) i t = σ(w i * [h t-1 , u t + b i ) o t = σ(w o * [h t-1 , u t + b o ) In Equation (13), f t , i t , c t represent the forget gate, input gate, and output gate respectively, h t represents the t-th output, h t-1 represents the (t - 1)-th output, c t represents the cell state, represents the candidate cell state, w and b represent the weight vector and bias vector, and σ and tanh represent the activation functions; The constructed LSTM neural network has a total of five layers. The first layer is the first LSTM layer, with 200 neurons set; the second layer is the Dropout layer, with the parameter set to 0.4; the third layer is the second LSTM layer, with 240 neurons set; the fourth layer is the Dropout layer, with the parameter set to 0.4; the last layer is the Dense layer, with the parameter set to 1; Use a for loop to take the 1st to 60th data of each subsequence as the first group of inputs, and the 61st data as the label of the first group of data; take the 2nd to 61st data as the second group of inputs, and the 62nd data as the label of the second group of data; and so on, until the (T - 60)th to (T - 1)th data are used as inputs, and the Tth input is used as the label of the last group of data. When the loop ends, the number of iterations of the model is set to 200; Through the above parameter settings, train the LSTM model. Input the primary subsequence u1(t), and automatically obtain the prediction result u1'(t) of the primary subsequence u1(t); ② Refer to step ① above to realize the prediction of the future short-term values of all primary subsequences and all secondary subsequences; ③ Sum the prediction results of all primary subsequences and secondary subsequences to obtain the prediction of the "future state" short-term power supply quantity of the target area, as shown in the following formula: In the above formula (14), u k '(t) represents the prediction result of the k-th primary subsequence, and e m '(t) represents the prediction result of the m-th secondary subsequence. GD t ' represents the prediction result of the power supply in the "future state" short-term target area.
4. A medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM according to claim 3, characterized in that, Specifically, the construction of the future electricity sales prediction model for the target area based on the collected historical electricity sales data of the target area is as follows. The construction steps are similar to those of the power supply prediction model, except for the different outputs of the models. The final output is the prediction result of the short-term electricity sales in the "future state" of the target area, denoted by SD t '.
5. A medium- and short-term line loss rate prediction method based on VMD-EWT-LSTM according to claim 4, characterized in that, The specific method for calculating the line loss rate of the target area in the future stage based on the predicted power supply quantity and power sales quantity of the target area at a future moment is as follows: Based on the predicted results of the power supply and power sales volume in the target area at a future moment, calculate the line loss rate in the target area in the future stage, denoted by FXS t It is represented by In the above formula, GD t ' represents the predicted result of the power supply quantity in the short-term target area of the "future state"; SD t ' represents the predicted result of the electricity sales quantity in the short-term target area of the "future state"; FXS t represents the predicted result of the line loss rate in the short-term target area of the "future state".
6. A medium- and short-term line loss rate prediction system based on VMD-EWT-LSTM, characterized in that, It includes, A data collection module for collecting historical power supply data and historical power sales data of a target area, represented by GD ij and SD ij respectively; The power supply quantity prediction model module is used to construct a power supply quantity prediction model for the target area at a future moment based on the historical power supply quantity data of the target area collected; The electricity sales quantity prediction model module is used to construct an electricity sales quantity prediction model for the target area at a future moment based on the historical electricity sales quantity data of the target area collected; The calculation module is used to predict the power supply quantity of the target area at a future moment based on the power supply quantity prediction model of the target area at a future moment, predict the electricity sales quantity of the target area at a future moment based on the electricity sales quantity prediction model of the target area at a future moment, and calculate the line loss rate of the target area in a future stage based on the predicted power supply quantity and electricity sales quantity of the target area at a future moment.
7. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the short-term and medium-term line loss rate prediction method based on VMD-EWT-LSTM as described in any one of claims 1-5.
8. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, it implements the steps of the short-term and medium-term line loss rate prediction method based on VMD-EWT-LSTM as described in any one of claims 1-5.