Load prediction method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a load prediction method and device, equipment and a storage medium. The method comprises the steps of firstly performing variational mode decomposition processing on node load data of a to-be-predicted node in a power system to obtain at least one load mode component, and then obtaining a load prediction result of the to-be-predicted node according to the at least one load mode component, a mode component strategy and a preset load prediction model, the modal component strategy is used for integrating or merging the at least one load modal component, and the load prediction model is determined based on the historical load data of the to-be-predicted node and the modal component strategy. According to the method provided by the invention, the load prediction efficiency is improved, and the accurate prediction of the load is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, and particularly relates to a load prediction method and device, equipment and a storage medium. BACKGROUND

[0002] With the transformation of energy structure and the increasing demand for intelligentization of power systems, intelligent and refined management of power systems is needed. Load prediction is crucial for stable operation and efficient management of power systems. In the new power system, the volatility, randomness and uncertainty of power load have significantly increased, especially with the access of distributed power and user-side comprehensive energy systems, making the load characteristics more complex and variable.

[0003] Existing load prediction methods are based on time series features extracted from time series of load data, and load prediction is performed by fitting the time series features.

[0004] However, the existing load prediction method has the problems of low load prediction efficiency and poor prediction accuracy. SUMMARY

[0005] The embodiments of the present application provide a load prediction method, device, equipment and storage medium to improve the efficiency and accuracy of load prediction.

[0006] In a first aspect, the embodiments of the present application provide a load prediction method, comprising:

[0007] Performing variational modal decomposition processing on node load data of a to-be-predicted node in a power system to obtain at least one load modal component;

[0008] Obtaining a load prediction result of the to-be-predicted node according to the at least one load modal component, a modal component strategy and a preset load prediction model, the modal component strategy being used for integrating or merging processing of the at least one load modal component, and the load prediction model being determined based on historical load data of the to-be-predicted node and the modal component strategy.

[0009] In one or more embodiments, the obtaining of the load prediction result of the to-be-predicted node according to the at least one load modal component, the modal component strategy and the preset load prediction model comprises:

[0010] Processing the at least one load modal component according to the modal component strategy to obtain an integrated result or a merged result;

[0011] Inputting the integrated result or the merged result into the load prediction model to obtain the load prediction result of the to-be-predicted node.

[0012] In one or more embodiments, before the load prediction result of the to-be-predicted node is obtained according to the at least one load modal component, the modal component strategy, and the preset load prediction model, the method further comprises:

[0013] obtaining at least one historical load modal component corresponding to the historical load data;

[0014] determining a historical integration result or a historical merging result corresponding to the at least one historical load modal component;

[0015] training a long short-term memory (LSTM) model according to the historical integration result or the historical merging result and real load data corresponding to the historical load data, to obtain the load prediction model.

[0016] In one or more embodiments, the variational modal decomposition processing of the node load data of the to-be-predicted node in the power system comprises:

[0017] decomposing the node load data to obtain at least one sub-modal sequence;

[0018] determining a variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence;

[0019] determining the at least one load modal component according to the variational problem.

[0020] In one or more embodiments, the determining of the variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence comprises:

[0021] for each sub-modal sequence, obtaining an analytic signal and a spectrum of the sub-modal sequence through Hilbert transform;

[0022] determining a fundamental frequency band corresponding to a frequency of the sub-modal sequence according to the analytic signal, the spectrum, and a preset center frequency;

[0023] processing a bandwidth in the fundamental frequency band of each sub-modal sequence according to a preset constraint condition to obtain the variational problem, the constraint condition being that a sum of each sub-modal sequence is equal to the node load data.

[0024] In one or more embodiments, the determining of the at least one load modal component according to the variational problem comprises:

[0025] adopting a Lagrange multiplier and a penalty factor to convert and process the variational problem to obtain an unconstrained variational problem;

[0026] solving the non-constrained variational problem to obtain at least one load modal component decomposed.

[0027] In one or more embodiments, the method further comprises:

[0028] inputting the load prediction result into a preset quantile regression model to obtain a prediction result corresponding to at least one confidence interval.

[0029] In a second aspect, the embodiments of the present application provide a load prediction device, comprising:

[0030] a first processing module configured to perform variational modal decomposition processing on node load data of a to-be-predicted node in a power system to obtain at least one load modal component;

[0031] a second processing module configured to obtain a load prediction result of the to-be-predicted node according to the at least one load modal component, a modal component strategy, and a preset load prediction model, the modal component strategy being used for integration or merging processing of the at least one load modal component, and the load prediction model being determined based on historical load data of the to-be-predicted node and the modal component strategy.

[0032] In one or more embodiments, the second processing module is specifically configured to:

[0033] process the at least one load modal component according to the modal component strategy to obtain an integration result or a merging result;

[0034] input the integration result or the merging result into the load prediction model to obtain the load prediction result of the to-be-predicted node.

[0035] In one or more embodiments, before the second processing module obtains the load prediction result of the to-be-predicted node according to the at least one load modal component, the modal component strategy, and the preset load prediction model, the second processing module is further configured to:

[0036] obtain at least one historical load modal component corresponding to the historical load data;

[0037] determine a historical integration result or a historical merging result corresponding to the at least one historical load modal component;

[0038] train a long short-term memory (LSTM) model according to the historical integration result or the historical merging result and real load data corresponding to the historical load data to obtain the load prediction model.

[0039] In one or more embodiments, the first processing module is specifically configured to:

[0040] decompose the node load data to obtain at least one sub-modal sequence;

[0041] determine a variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence;

[0042] determine the at least one load modal component according to the variational problem.

[0043] In one or more embodiments, the first processing module is configured to determine a variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence, and specifically configured to:

[0044] obtain an analytic signal and a spectrum of the sub-modal sequence through Hilbert transform for each sub-modal sequence;

[0045] determine a fundamental frequency band corresponding to a frequency of the sub-modal sequence according to the analytic signal, the spectrum, and a preset center frequency;

[0046] process a bandwidth in the fundamental frequency band of each sub-modal sequence according to a preset constraint condition to obtain the variational problem, and the constraint condition is that a sum of each sub-modal sequence is equal to the node load data.

[0047] In one or more embodiments, the first processing module is configured to determine the at least one load modal component according to the variational problem, and specifically configured to:

[0048] convert the variational problem through a Lagrange multiplier and a penalty factor to obtain an unconstrained variational problem;

[0049] solve the unconstrained variational problem to obtain the at least one decomposed load modal component.

[0050] In one or more embodiments, the second processing module is further configured to:

[0051] input the load prediction result into a preset quantile regression model to obtain a prediction result corresponding to at least one confidence interval.

[0052] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0053] The memory stores computer execution instructions.

[0054] The processor executes the computer execution instructions stored in the memory, so that the processor executes the method in the first aspect and any one of the embodiments.

[0055] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the method in the first aspect and any one of the embodiments.

[0056] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which is executed by a processor to implement the load prediction method in the first aspect and various possible implementation manners of the first aspect.

[0057] The load prediction method, device, equipment and storage medium provided by the embodiments of the present application. The method first performs variational modal decomposition processing on the node load data of the to-be-predicted node in the power system to obtain at least one load modal component, and then obtains the load prediction result of the to-be-predicted node according to the at least one load modal component, the modal component strategy, and the preset load prediction model, wherein the modal component strategy is used for integrating or merging processing of the at least one load modal component, and the load prediction model is determined based on the historical load data of the to-be-predicted node and the modal component strategy. In the above method, at least one load modal component is obtained by performing variational modal decomposition processing on the node load data of the to-be-predicted node in the power system, which effectively decomposes the node load data into load modal components with specific frequency components, can extract the node load data features in different frequency ranges, and provides a reliable basis for the accuracy of subsequent load prediction; by integrating or merging processing of the at least one load modal component according to the modal component strategy, and combining the preset load prediction model, the load prediction result of the to-be-predicted node can be obtained, which can effectively improve the accuracy and efficiency of load prediction. BRIEF DESCRIPTION OF DRAWINGS

[0058] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate one embodiment consistent with the present application and, together with the description, serve to explain the principles of the application.

[0059] Figure 1 Flowchart of the load prediction method provided by the embodiments of the present application Figure 1 ;

[0060] Figure 2 Flowchart of the load prediction method provided by the embodiments of the present application Figure 2 ;

[0061] Figure 3 Flowchart of the load prediction method provided by the embodiments of the present application Figure 3 ;

[0062] Figure 4 Structure diagram of the load prediction device provided by the embodiments of the present application;

[0063] Figure 5 A structural schematic diagram of an electronic device provided by an embodiment of the present application is shown.

[0064] The specific embodiments of the present application have been shown by the above-mentioned drawings, and will be described in more detail hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present application by any means, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0065] The exemplary embodiments will be described in detail herein with reference to the accompanying drawings. In the following description, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments are not representative of all embodiments consistent with the present application. Rather, they are merely examples of devices and methods consistent with some aspects of the present application, as detailed in the appended claims.

[0066] Before introducing the embodiments of the present application, the terms related to the embodiments of the present application are first explained:

[0067] Variational Mode Decomposition (VMD): refers to an adaptive, completely non-recursive mode variation and signal processing method, which realizes effective separation of intrinsic mode components, and determines the frequency center and bandwidth of each component by iteratively searching for the optimal solution of the variation model;

[0068] Long Short-Term Memory (LSTM) model: refers to a special neural network model that can effectively solve the problems of gradient disappearance and gradient explosion in traditional neural network models when processing long sequence data;

[0069] Hilbert transform: refers to a linear operator that generates a function with the same domain as the function, which can expand the time domain signal to the frequency domain.

[0070] Secondly, the application background of the embodiments of the present application is explained:

[0071] With the transformation of energy structure and the increasing demand for intelligentization of power systems, it is necessary to manage the power system intelligently and finely. Load forecasting is crucial for the stable operation and efficient management of power systems. In the new power system, the volatility, randomness and uncertainty of power load have significantly increased, especially with the integration of distributed power sources and user-side comprehensive energy systems, making the load characteristics more complex and variable.

[0072] The existing load prediction method extracts time sequence characteristics based on time sequence of load data, and performs load prediction by fitting the time sequence characteristics.

[0073] However, the existing load prediction method has problems of low load prediction efficiency and poor prediction accuracy.

[0074] The load prediction method provided by the present application aims to solve the above technical problems of the prior art. The inventive concept of the present application is as follows: the existing load prediction method has the problem of poor load prediction accuracy when fitting based on the extraction of time sequence characteristics from load data. Load prediction is mainly performed by extracting features from the time sequence of load data, and by analyzing the long-term dependence relationship between the extracted features, the load can be predicted. If a load prediction model that can capture the long-term dependence relationship of load sequence data can be constructed, the load can be quickly and accurately predicted. Therefore, the present application first performs variational modal decomposition processing on the node load data of the to-be-predicted node in the power system to obtain at least one load modal component. The complex node load data can be decomposed into simple load modal components, and each load modal component can represent the characteristics of the node load data. A modal component strategy can be set for the load modal component for integrating or merging at least one load modal component, and then a load prediction result of the to-be-predicted node is obtained according to the at least one load modal component, the modal component strategy, and a preset load prediction model.

[0075] The execution subject of the embodiment of the present application is an electronic device, which can be a terminal device such as a notebook computer, a desktop computer, a tablet computer, etc., and can also be a server. In actual application, whether the electronic device is a terminal device or a server can be determined according to actual conditions, and no specific limitation is made thereto.

[0076] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0077] Figure 1 Flowchart of the load prediction method provided by the embodiment of the present application Figure 1 As shown in Figure 1 , the load prediction method comprises the following steps:

[0078] S110, performing variational modal decomposition processing on the node load data of the to-be-predicted node in the power system to obtain at least one load modal component.

[0079] In this step, the node load data of the to-be-predicted node can be obtained from the power system, and the node load data of the to-be-predicted node is subjected to variational modal decomposition processing to obtain at least one load modal component.

[0080] For example, the node load data includes the load data of the to-be-predicted node in the power system within a period of time, that is, the node load data is time series data used to indicate the node load.

[0081] In a possible implementation, the variational modal decomposition processing on the node load data of the to-be-predicted node in the power system can be based on the node load data to construct a variational problem, and by solving the variational problem, the node load data is converted into a non-recursive variational modal decomposition form, that is, at least one load modal component, which can reflect different change rules in the node load data.

[0082] S120, obtaining a load prediction result of the to-be-predicted node according to the at least one load modal component, the modal component strategy, and a preset load prediction model;

[0083] The modal component strategy is used to integrate or merge the at least one load modal component, and the load prediction model is determined based on the historical load data of the to-be-predicted node and the modal component strategy.

[0084] For example, according to the at least one load modal component and the modal component strategy, the at least one load modal component is integrated or merged, and the processed at least one load modal component is taken as an input of the preset load prediction model, and the load prediction result of the to-be-predicted node is output by the load prediction model.

[0085] In a possible implementation, the above step S120 can further include the following steps:

[0086] Step 1, processing the at least one load modal component according to the modal component strategy to obtain an integration result or a merging result.

[0087] For example, according to the modal component strategy, the at least one load modal component is processed to obtain an integration result, which can be taking the at least one load modal component as a whole, and the obtained integration result is the whole composed of the at least one load modal component.

[0088] According to the modal component strategy, the at least one load modal component is processed to obtain a merging result, which can be performing correlation coefficient analysis on the at least one load modal component, removing high-frequency noise components in the at least one load modal component, and taking the obtained low-frequency modal component as a whole, and the obtained merging result is the at least one load modal component after removing the high-frequency noise components.

[0089] Step 2, input the integration result or the combination result into the load prediction model to obtain the load prediction result of the node to be predicted.

[0090] For example, the integration result or the combination result is input into the load prediction model, the load prediction model learns the correlation between the load modal components in the integration result or the combination result, and the features of each load modal component are extracted, the correlation between the features of each load modal component and the load prediction is learned, and the load prediction result of the node to be predicted is obtained.

[0091] In a possible implementation, at least one load modal component can be input into the load prediction model respectively to obtain at least one sub-load prediction result, and the at least one sub-load prediction result is superimposed to obtain the load prediction result of the node to be predicted.

[0092] The load prediction method provided by the embodiment of the present application first performs variational modal decomposition processing on the node load data of the node to be predicted in the power system to obtain at least one load modal component, and then obtains the load prediction result of the node to be predicted according to the at least one load modal component, the modal component strategy, and the preset load prediction model, wherein the modal component strategy is used for integration or combination processing of the at least one load modal component, and the load prediction model is determined based on the historical load data of the node to be predicted and the modal component strategy. In the embodiment, at least one load modal component is obtained by performing variational modal decomposition processing on the node load data of the node to be predicted in the power system, which effectively decomposes the node load data into load modal components with specific frequency components, can extract the node load data features in different frequency ranges, and provides a reliable basis for the accuracy of subsequent load prediction; the load prediction result of the node to be predicted is obtained by integrating or combining at least one load modal component according to the modal component strategy and combining the preset load prediction model, which can effectively improve the accuracy and efficiency of load prediction.

[0093] On the basis of the above embodiment, Figure 2 The flowchart of the load prediction method provided by the embodiment of the present application is shown in Figure 2 . As Figure 3 shown, before the above step S120, the load prediction method further includes the following steps:

[0094] S210, obtaining at least one historical load modal component corresponding to the historical load data.

[0095] In this step, the historical load data is processed by variational modal decomposition to obtain at least one historical load modal component. In order to train the load prediction model, at least one historical load modal component corresponding to the historical load data needs to be obtained.

[0096] Exemplarily, the at least one historical load modal component corresponding to the historical load data is divided into a model training set and a model verification set, the load prediction model is trained based on the model training set, and the accuracy of the trained load prediction model in predicting the load is analyzed based on the model verification set.

[0097] S220, determine a historical integration result or a historical merging result corresponding to the at least one historical load modal component.

[0098] In this step, according to the at least one historical load modal component, integration or merging processing can be performed to obtain a historical integration result or a historical merging result corresponding to the at least one historical load modal component.

[0099] In one possible implementation, the historical integration result corresponding to the at least one historical load modal component can be obtained by taking the at least one historical load modal component as a whole.

[0100] The historical merging result corresponding to the at least one historical load modal component can be obtained by performing correlation coefficient analysis on the at least one load modal component, removing high-frequency noise components in the at least one load modal component, and taking the obtained low-frequency modal component as a whole.

[0101] S230, training a long short-term memory network LSTM model according to the historical integration result or the historical merging result and real load data corresponding to the historical load data, to obtain a load prediction model.

[0102] In this step, the historical integration result or the historical merging result is input into the long short-term memory network LSTM model, the long short-term memory network LSTM model is trained according to load prediction results output by the model and real load data corresponding to the historical load data, and finally a load prediction model is obtained.

[0103] Exemplarily, the historical integration result or the historical merging result is a kind of sequence data, and the long short-term memory network LSTM model introduces memory cells and gating mechanisms, which can effectively capture long-term dependencies in sequence data.

[0104] The memory cell is mainly used for information transmission in the long short-term memory network LSTM model, and in the initial stage, the memory cell in the long short-term memory network LSTM model can be set to a zero vector.

[0105] The gating mechanism mainly includes a forgetting gate, an input gate, and an output gate.

[0106] Specifically, the forget gate can determine information that needs to be retained in the memory cell and information that needs to be deleted, according to input information of a current long short-term memory network (LSTM) model and a hidden state of a previous time (i.e., an output result of the previous time), and can determine the forget gate f t :

[0107] f t = σ (W f h t-1 + U f x t + b f )

[0108] wherein x t represents input information at a current t time, h t-1 represents a hidden state of a previous time (i.e., an output result of the previous time), W f and U f respectively represent a weight matrix of the forget gate, b f represents a bias term of the forget gate, and σ represents a sigmoid activation function.

[0109] The input gate can determine information that needs to enter the updated memory cell based on information retained in the memory cell, and the input gate i t may be represented as follows:

[0110] i t = σ (W i h t-1 + U i x t + b i )

[0111] wherein W i and U i respectively represent a weight matrix of the input gate, and b i represents a bias term of the input gate.

[0112] The output gate can determine information that needs to be output as a hidden state of a current time based on information retained in the memory cell, and the output gate o t may be represented as follows:

[0113] o t = σ (W o h t-1 + U o x t + b o )

[0114] wherein W o and U o respectively represent a weight matrix of the output gate, and b o represents a bias term of the output gate.

[0115] In a possible implementation, the candidate memory cell at the current moment can be determined according to input information of a current long short-term memory network (LSTM) model and a hidden state at a previous moment (i.e., an output result at the previous moment) may be represented as follows:

[0116]

[0117] wherein W c and U c respectively represent a weight matrix of the candidate memory cell, b c represents a bias term of the candidate memory cell.

[0118] According to the memory cell at the previous moment and the candidate memory cell at the current moment, the memory cell c t at the current moment can be calculated through a forgetting gate and an input gate, and can be represented as follows:

[0119]

[0120] wherein c t represents the memory cell at the current moment, c t-1 represents the memory cell at the previous moment, represents the candidate memory cell at the current moment, f t represents the forgetting gate, and i t represents the input gate.

[0121] According to the memory cell c t at the current moment, an output gate, and an activation function, the hidden state h t at the current moment (i.e., an output result at the current moment) can be determined, and can be represented as follows:

[0122] h t = o t ⊙ tanh(c t )

[0123] wherein h t represents the hidden state at the current moment, c t represents the memory cell at the current moment, o t represents the output gate, and tanh represents the activation function.

[0124] In a possible implementation, the long short-term memory network (LSTM) model can be trained according to a load prediction result output by the model and real load data corresponding to historical load data, and the mean square error and the root mean square error between the load prediction result output by the model and the real load data corresponding to the historical load data can be calculated as a training basis.

[0125] The mean square error is the mean of the sum of squares of the difference between the load prediction result output by the model and the real load data corresponding to the historical load data, and the mean square error can be expressed as:

[0126]

[0127] wherein e MSE represents the mean square error, Y true represents the real load data, Y pred represents the load prediction result, and M represents the number of load modal components decomposed in the load data. The smaller the value of the mean square error, the higher the fitting degree of the long short-term memory network LSTM model and the more accurate the prediction result.

[0128] The root mean square error is the square root of the ratio of the square of the difference between the load prediction result output by the model and the real load data corresponding to the historical load data to the sample number, and the root mean square error can be expressed as:

[0129]

[0130] wherein e RMSE represents the root mean square error, and the smaller the value of the root mean square error, the more accurate the prediction result of the long short-term memory network LSTM model.

[0131] In one possible implementation, the load prediction method further comprises: inputting the load prediction result into a preset quantile regression model to obtain a prediction result corresponding to at least one confidence interval.

[0132] For example, the quantile regression model comprises an input layer, a hidden layer, and an output layer, and the quantile regression formula can be expressed as:

[0133] Q Ytrue (τ|Y pred )=f(Y pred ,W(τ),V(τ))

[0134] wherein Y pred represents the load prediction result, τ represents the regression quantile, W(τ) represents the connection weight between the input layer and the hidden layer, and V(τ) represents the connection weight between the hidden layer and the output layer.

[0135] According to the estimated value of the connection weight W(τ) between the input layer and the hidden layer and the bias with the quantile condition The conditional quantile estimation formula is as follows:

[0136]

[0137] wherein, an estimate of the connection weight W(τ) between the input layer and the hidden layer, denotes the bias with quantile condition.

[0138] According to the conditional quantile estimation formula and the distribution function, the conditional probability density prediction can be expressed as:

[0139]

[0140] wherein, denotes the conditional probability density prediction corresponding to the conditional quantile estimation, the point prediction value can be converted into a prediction result corresponding to a confidence interval, such as a quantile point τ1=0.1 corresponding to a 10% quantile and a quantile point τ2=0.9 corresponding to a 90% quantile.

[0141] The load prediction method provided by the embodiment of the present application first acquires at least one historical load modal component corresponding to historical load data, then determines a historical integration result or a historical combination result corresponding to the at least one historical load modal component, and finally trains a long short-term memory network LSTM model according to the historical integration result or the historical combination result and real load data corresponding to the historical load data, to obtain a load prediction model. In the embodiment, at least one historical load modal component corresponding to historical load data is acquired, and the at least one historical load modal component is processed to obtain a corresponding historical integration result or a historical combination result, so that the input of the long short-term memory network LSTM model can be determined based on the historical load data. By inputting the historical integration result or the historical combination result into the long short-term memory network LSTM model, the output can obtain a prediction result of the historical load data. By comparing the prediction result with real load data corresponding to the historical load data, the long short-term memory network LSTM model is trained to obtain a load prediction model, which can effectively improve the prediction accuracy of the load prediction model and realize accurate load prediction.

[0142] On the basis of the above embodiment, Figure 3 The flowchart of the load prediction method provided by the embodiment of the present application is shown in Figure 4 . As Figure 4 shown, one possible implementation of the above step S110 further includes the following steps:

[0143] S310, decompose the node load data to obtain at least one sub-modal sequence.

[0144] Illustratively, the sub-modal sequence refers to a sequence with a preset center frequency and a limited bandwidth.

[0145] In one possible implementation, at least one sub-modal sequence u k (t) obtained by decomposing the node load data x(t) can be expressed as follows:

[0146] u k (t)=A k (t)cos(φ k (t))

[0147] Where k represents the number of submodal sequences, A k (t) is the submode sequence u k The envelope amplitude of (t), φ k (t) is the submode sequence u k The phase of (t), and φ k (t) is a non-decreasing function, i.e., φ k The first derivative φ′ of (t) k (t)≥0.

[0148] S320. Based on at least one submode sequence, determine the variational problem corresponding to the variational mode decomposition process.

[0149] For example, the bandwidth corresponding to at least one sub-mode sequence is obtained, and the variational problem corresponding to the variational mode decomposition process is determined based on the minimum sum of the bandwidths corresponding to at least one sub-mode sequence.

[0150] In one possible implementation, step S320 may further include the following steps:

[0151] Step 1: For each sub-mode sequence, obtain the analytic signal and spectrum of the sub-mode sequence through Hilbert transform.

[0152] For example, for each sub-mode sequence, for the sub-mode sequence u k The analytic signal and spectrum of the submode sequence obtained by performing a Hilbert transform on (t) can be represented as follows:

[0153]

[0154] Where δ(t) represents the Dirac distribution, and * represents the convolution symbol.

[0155] Step 2: Determine the fundamental frequency band corresponding to the frequency of the sub-mode sequence based on the analytical signal, spectrum, and preset center frequency.

[0156] In one possible implementation, the frequency of the submode sequence can be transformed to the correct fundamental frequency band, i.e., the fundamental frequency band corresponding to the frequency of the submode sequence, based on a preset center frequency.

[0157] For example, the preset center frequency is The submode sequence u is determined by multiplying the analytical signal and its spectrum by a preset center frequency. kThe base frequency band corresponding to the frequency of (t) can be expressed as follows:

[0158]

[0159] wherein δ(t) represents a Dirac distribution, and * represents a convolution symbol, ω0 represents a preset center frequency.

[0160] Step 3, according to the preset constraint condition, processing the bandwidth in the base frequency band of each sub-modal sequence to obtain a variational problem.

[0161] wherein the constraint condition is that the sum of each sub-modal sequence is equal to the node load data.

[0162] In a possible implementation, an L2 norm can be added to the bandwidth in the base frequency band of each sub-modal sequence for regularization to prevent overfitting.

[0163] For example, according to the constraint condition that the sum of each sub-modal sequence is equal to the node load data, the sum of the bandwidth in the base frequency band of each sub-modal sequence is minimized to obtain the variational problem, which can be expressed as follows:

[0164]

[0165] wherein k represents the number of sub-modal sequences, {u k} represents a set of sub-modal sequences, that is, {u k}={u1,u2,...,u K}, {ω k} represents a set of preset center frequencies, that is, {ω k}={ω1,ω2,...,ω K}, and x(t) represents the node load data.

[0166] S330, determining at least one load modal component according to the variational problem.

[0167] For example, solving the variational problem can determine at least one load modal component.

[0168] In a possible implementation, the above step S330 can further include the following steps:

[0169] Step 1, using a Lagrange multiplier and a penalty factor to convert and process the variational problem to obtain an unconstrained variational problem.

[0170] For example, in order to solve the variational problem, a Lagrange multiplier and a penalty factor can be used to convert the variational problem into an unconstrained variational problem.

[0171] The unconstrained variational problem can be expressed as follows:

[0172]

[0173] where λ(t) represents the Lagrange multiplier, and α represents the penalty factor.

[0174] Step 2, solving the unconstrained variational problem to obtain the decomposed at least one load modal component.

[0175] In one possible implementation, the unconstrained variational problem can be solved by using a multiplier alternating direction method, by alternately updating the optimal value of the unconstrained variational problem, i.e., the modal component, the center frequency corresponding to the modal component, and the Lagrange multiplier, and stopping the updating when a preset stopping condition is met, to obtain the decomposed at least one load modal component.

[0176] For example, the update of the modal component can be expressed as follows:

[0177]

[0178] where n represents the iteration number, represents the Fourier transform of the updated modal component u k represents the Fourier transform of the node load data x(t), represents the Fourier transform of the sub-modal sequence u i represents the Fourier transform of the Lagrange multiplier λ(t).

[0179] The update of the center frequency corresponding to the modal component can be expressed as follows:

[0180]

[0181] where, represents the Fourier transform of the sub-modal sequence u k

[0182] The preset stopping condition can be expressed as follows:

[0183] In one possible implementation, a grid search strategy can be used to optimize the key parameters in the variational modal decomposition process, which can include the number k of the decomposed sub-modal sequence and the penalty factor α.

[0184] ​​​Exemplarily, candidate ranges of the respective key parameters are set in advance according to characteristics of the node load data, a grid space composed of combinations of the respective key parameters is generated, a modal component decomposition quality score corresponding to each combination of the key parameters is respectively calculated, and the optimal combination of the key parameters is selected as the parameter of the variational modal decomposition.

[0185] The load prediction method provided in the embodiments of the present application first decomposes the node load data to obtain at least one sub-modal sequence, then determines a variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence, and then determines at least one load modal component according to the variational problem. In the embodiments, the complex load data can be decomposed to obtain at least one sub-modal sequence through the decomposition of the node load data, and different types of feature information in the node load data can be effectively extracted; the variational problem corresponding to the variational modal decomposition processing is determined according to the at least one sub-modal sequence, and then the variational problem is solved, so that at least one load modal component can be determined, the feature information of the data can be better understood and utilized, and the accuracy of the load prediction can be improved.

[0186] On the basis of the above-mentioned embodiments, the load prediction device provided in the embodiments of the present application can perform the method provided in the method embodiments.

[0187] Figure 5 The structure diagram of the load prediction device provided in the embodiments of the present application is shown in FIG. 4. As shown in FIG. 4, the load prediction device 400 comprises: Figure 5

[0188] The first processing module 410 is configured to perform variational modal decomposition processing on the node load data of the node to be predicted in the power system to obtain at least one load modal component.

[0189] The second processing module 420 is configured to obtain a load prediction result of the node to be predicted according to the at least one load modal component, a modal component strategy, and a preset load prediction model, wherein the modal component strategy is used for integration or merging processing of the at least one load modal component, and the load prediction model is determined based on historical load data of the node to be predicted and the modal component strategy.

[0190] In one or more embodiments, the second processing module 420 is specifically configured to:

[0191] process the at least one load modal component according to the modal component strategy to obtain an integration result or a merging result;

[0192] input the integration result or the merging result into the load prediction model to obtain the load prediction result of the node to be predicted.

[0193] ​In one or more embodiments, before obtaining the load prediction result of the to-be-predicted node according to the at least one load modal component, the modal component strategy, and the preset load prediction model, the second processing module 420 is further configured to:

[0194] obtain at least one historical load modal component corresponding to the historical load data;

[0195] determine a historical integration result or a historical merging result corresponding to the at least one historical load modal component;

[0196] train a long short-term memory network (LSTM) model according to the historical integration result or the historical merging result and real load data corresponding to the historical load data, to obtain the load prediction model.

[0197] In one or more embodiments, the first processing module 410 is specifically configured to:

[0198] decompose the node load data to obtain at least one sub-modal sequence;

[0199] determine a variational problem corresponding to a variational modal decomposition process according to the at least one sub-modal sequence;

[0200] determine the at least one load modal component according to the variational problem.

[0201] In one or more embodiments, the first processing module 410 determines the variational problem corresponding to the variational modal decomposition process according to the at least one sub-modal sequence, and is specifically configured to:

[0202] obtain an analytic signal and a spectrum of each sub-modal sequence through Hilbert transform;

[0203] determine a fundamental frequency band corresponding to a frequency of the sub-modal sequence according to the analytic signal, the spectrum, and a preset center frequency;

[0204] process a bandwidth in the fundamental frequency band of each sub-modal sequence according to a preset constraint condition to obtain the variational problem, wherein the constraint condition is that a sum of each sub-modal sequence is equal to the node load data.

[0205] In one or more embodiments, the first processing module 410 determines the at least one load modal component according to the variational problem, and is specifically configured to:

[0206] convert the variational problem through a Lagrange multiplier and a penalty factor to obtain an unconstrained variational problem;

[0207] solve the unconstrained variational problem to obtain the at least one decomposed load modal component.

[0208] In one or more embodiments, the second processing module 420 is further configured to:

[0209] inputting the load prediction result into a preset quantile regression model to obtain a prediction result corresponding to at least one confidence interval.

[0210] On the basis of the above-mentioned embodiments, ​ A structural schematic diagram of an electronic device is provided in the embodiments of the present application. As shown in the structural schematic diagram, ​ The electronic device 500 includes a processor 510, a memory 520 and a bus 530.

[0211] The memory 520 is configured to store computer-executed instructions of the processor 510.

[0212] The processor 510 is configured to execute the technical solutions of any one of the method embodiments by executing the computer-executed instructions.

[0213] Optionally, the memory 520 can be independent or integrated with the processor 510.

[0214] Optionally, the memory 520 can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory.

[0215] The bus 530 can be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the drawings of the present application, but it does not mean that there is only one bus or only one type of bus.

[0216] The processor described above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0217] The electronic device is configured to execute the technical solutions of any one of the method embodiments, and the implementation principles and technical effects are similar, which will not be described here.

[0218] The embodiment of the present application further provides a computer readable storage medium, which stores computer execution instructions, and the computer execution instructions are used for realizing the technical solutions provided by any of the method embodiments when executed by a processor.

[0219] The embodiment of the present application further provides a computer program product, which comprises a computer program, and the computer program product comprises computer instructions stored in a computer readable storage medium, and the computer program is used for realizing the technical solutions provided by any of the method embodiments when executed by a processor.

[0220] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited to the action order described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0221] It should be further noted that, although each step in the flowchart is displayed in sequence according to the arrow indication, these steps are not necessarily executed in sequence according to the arrow indication. Unless explicitly stated in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or sub-steps or stages of other steps.

[0222] It should be understood that the above-mentioned device embodiments are only schematic, and the device of the present application can also be realized by other ways. For example, the division of units / modules in the above-mentioned embodiments is only a logical function division, and another division way can be used in actual implementation. For example, multiple units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0223] In addition, each functional unit / module in each embodiment of the present application can be integrated in one unit / module, or each unit / module can exist physically, or two or more units / modules can be integrated together. The integrated unit / module can be realized in the form of hardware or in the form of software program module.

[0224] If the integrated units / modules are implemented in the form of hardware, the hardware can be a digital circuit, an analog circuit, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any appropriate hardware processor, such as a CPU, a GPU, an FPGA, a DSP, an ASIC, etc. Unless otherwise specified, the storage unit can be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc.

[0225] If the integrated units / modules are implemented in the form of software program modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the essential part or all or part of the technical solutions that make contributions to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0226] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0227] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope of the application being indicated by the appended claims.

Claims

1. A load prediction method characterized by, The method comprises: performing variational modal decomposition processing on node load data of a node to be predicted in a power system to obtain at least one load modal component; obtaining a load prediction result of the node to be predicted according to the at least one load modal component, a modal component strategy and a preset load prediction model, wherein the modal component strategy is used for integrating or merging processing of the at least one load modal component, and the load prediction model is determined based on historical load data of the node to be predicted and the modal component strategy.

2. The method of claim 1, wherein, The load prediction result of the node to be predicted is obtained according to the at least one load modal component, the modal component strategy and the preset load prediction model, comprising: processing the at least one load modal component according to the modal component strategy to obtain an integrated result or a merged result; inputting the integrated result or the merged result into the load prediction model to obtain the load prediction result of the node to be predicted.

3. The method of claim 1, wherein, Before the load prediction result of the node to be predicted is obtained according to the at least one load modal component, the modal component strategy and the preset load prediction model, the method further comprises: obtaining at least one historical load modal component corresponding to the historical load data; determining a historical integrated result or a historical merged result corresponding to the at least one historical load modal component; training a long short-term memory (LSTM) model according to the historical integrated result or the historical merged result and real load data corresponding to the historical load data to obtain the load prediction model.

4. The method according to any one of claims 1 to 3, characterized in that, The at least one load modal component is obtained by performing variational modal decomposition processing on node load data of a node to be predicted in a power system, comprising: decomposing the node load data to obtain at least one sub-modal sequence; determining a variational problem corresponding to the variational modal decomposition processing according to the at least one sub-modal sequence; determining the at least one load modal component according to the variational problem.

5. The method of claim 4, wherein, The variational problem corresponding to the variational modal decomposition processing is determined according to the at least one sub-modal sequence, comprising: for each sub-modal sequence, obtaining an analytic signal and a frequency spectrum of the sub-modal sequence through Hilbert transform; determining a fundamental frequency band corresponding to a frequency of the sub-modal sequence according to the analytic signal, the frequency spectrum and a preset center frequency; processing a bandwidth in the fundamental frequency band of each sub-modal sequence according to a preset constraint condition to obtain the variational problem, wherein the constraint condition is that a sum of each sub-modal sequence is equal to the node load data.

6. The method of claim 4, wherein, The at least one load modal component is determined according to the variational problem, comprising: performing conversion processing on the variational problem by using a Lagrange multiplier and a penalty factor to obtain an unconstrained variational problem; solving the unconstrained variational problem to obtain at least one decomposed load modal component.

7. The method according to any one of claims 1 to 3, characterized in that, The method further comprises: inputting the load prediction result into a preset quantile regression model to obtain a prediction result corresponding to at least one confidence interval.

8. A load prediction device characterized by comprising: comprising: The first processing module is configured to perform variational mode decomposition on node load data of a node to be predicted in a power system to obtain at least one load mode component. The second processing module is configured to obtain a load prediction result of the node to be predicted according to the at least one load mode component, a mode component strategy, and a preset load prediction model, the mode component strategy being used for integration or combination of the at least one load mode component, and the load prediction model being determined based on historical load data of the node to be predicted and the mode component strategy.

9. An electronic device, comprising: The method comprises: a memory and a processor; the memory stores computer-executed instructions; the processor executes the computer-executed instructions stored in the memory, so that the processor executes the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executed instructions, and the computer-executed instructions are executed by the processor to implement the method according to any one of claims 1-7.

Images