CN111126659A - Power load prediction method and system - Google Patents

Power load prediction method and system Download PDF

Info

Publication number
CN111126659A
CN111126659A CN201911128175.XA CN201911128175A CN111126659A CN 111126659 A CN111126659 A CN 111126659A CN 201911128175 A CN201911128175 A CN 201911128175A CN 111126659 A CN111126659 A CN 111126659A
Authority
CN
China
Prior art keywords
component
sequence
load prediction
load
modal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201911128175.XA
Other languages
Chinese (zh)
Inventor
孙正来
马骏
丁倩
徐璐
江涛
余述良
徐斌
李葆
汤远红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Liuan Power Supply Co of State Grid Anhui Electric Power Co Ltd
Original Assignee
Liuan Power Supply Co of State Grid Anhui Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Liuan Power Supply Co of State Grid Anhui Electric Power Co Ltd filed Critical Liuan Power Supply Co of State Grid Anhui Electric Power Co Ltd
Priority to CN201911128175.XA priority Critical patent/CN111126659A/en
Publication of CN111126659A publication Critical patent/CN111126659A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Primary Health Care (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention discloses a power load prediction method and a system, which belong to the technical field of power, and comprise the steps of decomposing an original load sequence by using a lumped empirical mode decomposition algorithm; calculating approximate entropy of each modal component and combining to obtain a reconstructed new sequence; each new subsequence is predicted by a load prediction model of the extreme learning machine; and superposing the prediction results of each subsequence to obtain a final prediction value. The prediction analysis of the actual power grid load data by using the method provided by the invention shows that the method effectively improves the prediction precision.

Description

Power load prediction method and system
Technical Field
The invention relates to the technical field of electric power, in particular to a method and a system for predicting an electric power load.
Background
At present, a plurality of load prediction methods exist, and the traditional prediction method mainly comprises a regression analysis method, a time series method, a trend extrapolation method and the like. The methods adopt mathematical ideas to establish corresponding prediction models, depend on historical data, and are difficult to solve the randomness of loads. Modern prediction technologies include fuzzy prediction, support vector machines, neural networks, and the like. For the neural network, a static Back Propagation (BP) is frequently applied, and although the BP has a certain self-learning capability, the convergence speed is slow, and the BP is easy to fall into a local minimum value, so that the application of the BP is limited. The neural network (ELM) is a newer neural network algorithm, can randomly select hidden node parameters, has good global search capability, and overcomes the defect of overfitting of the traditional neural network.
In order to further improve the short-term load prediction precision, the combined prediction model is widely applied to the aspect of load prediction. The method mainly comprises the steps of decomposing a target load sequence into a plurality of independent internal modes by an EMD (empirical mode decomposition) based decomposition method, and then respectively establishing a prediction model for each mode. However, the EMD decomposition method is difficult to avoid the occurrence of modal aliasing, and the obtained false IMF adversely affects the prediction accuracy.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and improve the load prediction precision.
In order to achieve the above object, the present invention adopts a power load prediction method, including the steps of:
decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components;
calculating the approximate entropy of each modal component, and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
respectively carrying out load prediction on each component sequence by using a load prediction model based on an extreme learning machine to obtain a load prediction result corresponding to each component sequence;
and superposing the load prediction results corresponding to each component sequence to obtain a load prediction value corresponding to the original load sequence.
Further, the calculating the approximate entropy of each modal component, and superimposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component includes:
performing m-dimensional reconstruction on the one-dimensional time sequence of each modal component to obtain vectors V (j) and V (q);
calculating the distance d [ V (j), V (q) ] of two vectors V (j), V (q);
according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure BDA0002277510900000021
Calculating the average autocorrelation degree of the vector V (j) according to the correlation integral;
calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j);
and overlapping the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
Further, the load prediction of each component sequence by using the load prediction model based on the extreme learning machine to obtain the load prediction result corresponding to each component sequence includes:
using the sample composed of the component sequences as an input vector of the extreme learning machine-based load prediction model, and using the expected output value corresponding to each component sequence as a load prediction result corresponding to each component sequence, wherein the load prediction model of the extreme learning machine is as follows:
Figure BDA0002277510900000022
wherein the content of the first and second substances,
Figure BDA0002277510900000023
g (x) is the activation function,
Figure BDA0002277510900000024
for the weights of the input layer to the hidden layer,
Figure BDA0002277510900000025
for the deviation of the i-th implicit node,
Figure BDA0002277510900000026
the weights for the hidden layer to the output layer,
Figure BDA0002277510900000027
for the sequence of components of the input the,
Figure BDA0002277510900000028
n is the number of samples for the desired output value.
On the other hand, the power load prediction system comprises a decomposition module, a component sequence calculation module, a load prediction module and a load superposition module;
the decomposition module is used for decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components;
the component sequence calculation module is used for calculating the approximate entropy of each modal component and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
the load prediction module is used for respectively predicting the load of each component sequence by using a load prediction model based on the extreme learning machine to obtain a load prediction result corresponding to each component sequence;
and the load superposition module is used for superposing the load prediction results corresponding to each component sequence to obtain the load prediction value corresponding to the original load sequence.
Further, the component sequence calculation module comprises a reconstruction unit, an inter-vector distance calculation unit, a correlation integral calculation unit, an average autocorrelation degree calculation unit, an approximate entropy calculation unit and a superposition unit;
the reconstruction unit is used for performing m-dimensional reconstruction on the one-dimensional time sequence of each modal component to obtain vectors V (j) and V (q);
the inter-vector distance calculation unit is used for calculating the distance d [ V (j), V (q) ] of two vectors V (j), V (q);
the correlation integral calculation unit is used for calculating the correlation integral according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure BDA0002277510900000031
The average autocorrelation degree calculating unit is used for calculating the average autocorrelation degree of the vector V (j) according to the correlation integral;
the approximate entropy calculation unit is used for calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j);
the superposition unit is used for superposing the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
Further, the load prediction module comprises a sample construction unit and a prediction unit;
the sample construction unit is used for constructing samples according to the component sequences;
the prediction unit is used for taking the samples as input vectors of the extreme learning machine-based load prediction model, and taking expected output values corresponding to each component sequence as load prediction results corresponding to each component sequence, wherein the load prediction model of the extreme learning machine is as follows:
Figure BDA0002277510900000041
wherein the content of the first and second substances,
Figure BDA0002277510900000042
in order to imply the number of layer nodes,g (x) is the activation function,
Figure BDA0002277510900000043
for the weights of the input layer to the hidden layer,
Figure BDA0002277510900000044
for the deviation of the i-th implicit node,
Figure BDA0002277510900000045
the weights for the hidden layer to the output layer,
Figure BDA0002277510900000046
for the sequence of components of the input the,
Figure BDA0002277510900000047
n is the number of samples for the desired output value.
Compared with the prior art, the invention has the following technical effects: according to the method, the original load sequence is decomposed by adopting a lumped empirical mode decomposition method, so that the mode aliasing problem caused by discontinuity of signals of the original load sequence is solved; meanwhile, the approximate entropy is applied to load decomposition, so that the operation scale is greatly reduced; and finally, respectively carrying out load prediction on each component sequence by using a load prediction model based on the extreme learning machine, and superposing the load prediction results to obtain a final load prediction value, so that the prediction of actual power grid load data is more accurate, and the effectiveness of the power grid load prediction result is improved.
Drawings
The following detailed description of embodiments of the invention refers to the accompanying drawings in which:
FIG. 1 is a flow diagram of a method of power load prediction;
FIG. 2 is a schematic diagram of power load prediction;
FIG. 3 is a schematic diagram of an electrical load prediction system;
FIG. 4 is a diagram illustrating the result of decomposition of an original load sequence using EEMD;
fig. 5 is a schematic diagram of a component sequence corresponding to each modal component;
fig. 6 is a schematic diagram of a load prediction result corresponding to an original load sequence.
Detailed Description
To further illustrate the features of the present invention, refer to the following detailed description of the invention and the accompanying drawings. The drawings are for reference and illustration purposes only and are not intended to limit the scope of the present disclosure.
As shown in fig. 1-2, the present embodiment discloses a power load prediction method, which includes the following steps S1 to S4:
s1, decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components;
s2, calculating the approximate entropy of each modal component, and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
s3, respectively carrying out load prediction on each component sequence by using a load prediction model based on the extreme learning machine to obtain a load prediction result corresponding to each component sequence;
and S4, overlapping the load prediction results corresponding to each component sequence to obtain a load prediction value corresponding to the original load sequence.
It should be noted that, in this embodiment, an Empirical Mode Decomposition (EMD) algorithm may also be used to decompose the original load sequence, and EMD is a method for smoothing a nonlinear and non-stationary signal. The original sequence is decomposed into inherent modal functions under different scales through a series of screens, and the functions and components reflect local characteristic information of fluctuation or trend of the original signal under different scales. However, the EMD decomposition method is difficult to avoid the generation of modal aliasing, and the obtained false IMF will have adverse effect on the accuracy of the fitting prediction.
In order to solve the modal aliasing of EMD Decomposition caused by discontinuity of the original load sequence, the present embodiment adopts lumped Empirical Mode Decomposition (EEMD), introduces white noise into the signal to be analyzed in the Hilbert-Huang transformation process, and the frequency spectrum of the white noise is uniformly distributed, so that the signal can be automatically distributed on a proper reference scale, thereby well solving the modal aliasing problem.
Specifically, step S1 described above: decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components, wherein the specific steps are as follows S11-S14:
s11, the number of times K the EMD algorithm is executed is determined.
S12, the kth EMD decomposition is performed (K is 1,2, …, K).
S13, adding white noise signal l to the target signal x (t) before each decompositionk(t) obtaining a new signal xk(t); decomposition of x with EMDk(t) obtaining M components (IMF)i，kSubscripts denote the ith IMF from the kth experimental decomposition; if k is<K, K is K +1, and the above steps S12 to S13 are repeated.
It should be noted that the target signal x (t) refers to the original load when EMD is performed for the first time, and then refers to the newly obtained xk(t)。
And S14, calculating an average value of each IMF obtained by K times of decomposition, namely the final modal component IMF, wherein the IMF component with the lowest frequency and the smoothest frequency is the residual component RES).
Specifically, step S2 described above: calculating the approximate entropy of each modal component, and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component, specifically including the following steps S21 to S26:
s21, performing m-dimensional reconstruction on the one-dimensional time series of each modal component to obtain vectors v (j), v (q), where:
V(j)＝[v(j),v(j+1),v(j+2),...,v(j+m-1)]
where j is 1, 2.. P + m-1, and the length of the one-dimensional time series is P.
S22, calculating the distance d [ V (j), V (q) ] between two vectors V (j), V (q), and the distance between the two vectors V (j), V (q) is defined as:
Figure BDA0002277510900000061
s23, according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure BDA0002277510900000062
Integration of correlation
Figure BDA0002277510900000063
Comprises the following steps:
Figure BDA0002277510900000064
in the formula: θ (·) is the Heaviside function, b is the tolerance, and m is 2, and b is 0.15 σ.
S24, calculating the average autocorrelation degree of the vector V (j) according to the correlation integral, wherein the calculation formula is as follows:
Figure BDA0002277510900000071
s25, calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j), wherein the calculation formula is as follows:
β(m,b,P)＝αm(b)-αm+1(b)。
and S26, overlapping the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
Specifically, step S3 described above: respectively carrying out load prediction on each component sequence by using a load prediction model based on an extreme learning machine to obtain a load prediction result corresponding to each component sequence, and specifically comprising the following steps of S31-S32:
s31, forming a sample by the component sequence;
s32, using the sample as an input vector of the extreme learning machine-based load prediction model
Figure BDA0002277510900000072
Expected output values to be corresponding to each component sequence
Figure BDA0002277510900000073
As a load prediction result corresponding to each component sequence, the load prediction model of the extreme learning machine is:
Figure BDA0002277510900000074
wherein the content of the first and second substances,
Figure BDA0002277510900000075
g (x) is the activation function,
Figure BDA0002277510900000076
for the weights of the input layer to the hidden layer,
Figure BDA0002277510900000077
for the deviation of the i-th implicit node,
Figure BDA0002277510900000078
the weights for the hidden layer to the output layer,
Figure BDA0002277510900000079
for the sequence of components of the input the,
Figure BDA00022775109000000710
n is the number of samples for the desired output value.
Further, the load prediction model of the extreme learning machine is in the form of a written matrix as follows:
Hμ＝O
in the formula: h is the output of the hidden layer node, and the output weight mu is solved by the following least square solution:
Figure BDA0002277510900000081
as shown in fig. 3, the present embodiment discloses an electrical load prediction system, which includes a decomposition module 10, a component sequence calculation module 20, a load prediction module 30, and a load superposition module 40;
the decomposition module 10 is configured to decompose the original load sequence by using a lumped empirical mode decomposition method to obtain a plurality of modal components;
the component sequence calculation module 20 is configured to calculate an approximate entropy of each modal component, and superimpose the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
the load prediction module 30 is configured to perform load prediction on each component sequence by using a load prediction model based on an extreme learning machine, so as to obtain a load prediction result corresponding to each component sequence;
the load superposition module 40 is configured to superpose the load prediction results corresponding to each component sequence, so as to obtain a load prediction value corresponding to the original load sequence.
Specifically, the component sequence calculation module 20 includes a reconstruction unit, an inter-vector distance calculation unit, a correlation integral calculation unit, an average autocorrelation degree calculation unit, an approximate entropy calculation unit, and a superposition unit;
the reconstruction unit is used for performing m-dimensional reconstruction on the one-dimensional time sequence of each modal component to obtain vectors V (j) and V (q);
the inter-vector distance calculation unit is used for calculating the distance d [ V (j), V (q) ] of two vectors V (j), V (q);
the correlation integral calculation unit is used for calculating the correlation integral according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure BDA0002277510900000082
The average autocorrelation degree calculating unit is used for calculating the average autocorrelation degree of the vector V (j) according to the correlation integral;
the approximate entropy calculation unit is used for calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j);
the superposition unit is used for superposing the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
Specifically, the load prediction module 30 includes a sample construction unit and a prediction unit;
the sample construction unit is used for constructing samples according to the component sequences;
the prediction unit is used for taking the samples as input vectors of the extreme learning machine-based load prediction model, and taking expected output values corresponding to each component sequence as load prediction results corresponding to each component sequence, wherein the load prediction model of the extreme learning machine is as follows:
Figure BDA0002277510900000091
wherein the content of the first and second substances,
Figure BDA0002277510900000092
g (x) is the activation function,
Figure BDA0002277510900000093
for the weights of the input layer to the hidden layer,
Figure BDA0002277510900000094
for the deviation of the i-th implicit node,
Figure BDA0002277510900000095
the weights for the hidden layer to the output layer,
Figure BDA0002277510900000096
for the sequence of components of the input the,
Figure BDA0002277510900000097
n is the number of samples for the desired output value.
The effect of the power load prediction scheme designed in the embodiment is verified through experimental simulation and analysis as follows:
load prediction is performed on a certain city of Anhui province, 960 pieces of real-time measurement data from No. 21 to No. 30 in 2018 are selected in the embodiment, the data of the previous 8 days are used as training samples, the data of the last two days are used as test samples to perform simulation experiments, and the time sampling interval is 15min, namely 96 points per day.
(1) EEMD decomposition
Decomposition of the raw load data using EEMD is performed a number of times set to 100 and the noise variance is set to 0.3, resulting in the following fig. 4: 8 Intrinsic Mode Function (IMF) components and one component are obtained, the IMF component shows a change rule from high frequency to low frequency, the volatility is high, and the number of the components is large. If the prediction is carried out by directly adopting an ELM (extreme learning machine) conforming prediction model, the complexity of calculation is increased and the prediction effect is poor.
In this embodiment, by calculating the approximate entropy of each component (see table 1), the components with similar values are combined to obtain a new component, as shown in fig. 5. From table 1, the approximate entropy of each component decreases with decreasing frequency, indicating that the complexity decreases from high frequency to low frequency components, verifying the validity of the approximate entropy. The periods of the IMF1, IMF2, and IMF3 have a certain difference, approximate entropies are calculated for the three components, the obtained values are close, and therefore the three components are superposed. And performing complexity analysis on the rest components, and overlapping the components with approximate entropy values. Finally, three new subsequences are obtained as shown in FIG. 5. As can be seen from FIG. 5, the subsequences reconstructed by EEMD-sample entropy have regularity and can better represent load characteristics. F1 shows obvious randomness, reflecting that the load is influenced by random factors; f2 is a periodic component with obvious regularity; the F3 component fluctuates slowly and the curve is smooth.
TABLE 1 approximate entropy of the components
IMF component Approximate entropy
IMF1 1.4287
IMF2 0.96
IMF3 0.62
IMF4 0.46
IMF5 0.34
IMF6 0.17
IMF7 0.062
IMF8 0.02
RES 0.0073
(2) Analysis of predicted results
Based on the analysis, the reconstructed 3 subsequences are respectively subjected to load prediction by adopting ELM, 4 EEMD-approximate entropy-ELM combined models are established, and the results are superposed to obtain the final prediction result. Wherein, the hidden layer neuron defaults to 30, the activation function is sig, and the TYPE defaults to 0.
In order to verify the effectiveness of the EEMD-approximate entropy-ELM prediction model (marked as model one) provided by the embodiment, the EEMD-approximate entropy-BP algorithm (marked as model two) and the prediction results of the EMD-ELM algorithm are respectively compared (marked as model three). The predicted index uses the mean absolute percent error MAPE and the root mean square error MSE. The results of the prediction of the three models are shown in fig. 6, and it can be seen that the error of the EEMD-approximate entropy-BP prediction method and the EMD-ELM prediction method is large, and the accuracy of the EEMD-approximate entropy-ELM prediction method is obviously higher than that of the other two methods at several points of rapid load change.
Table 2 shows three model error indexes, and compared with the EEMD-approximate entropy-BP method, the EEMD-approximate entropy-ELM prediction method has the advantages that the average MAPE and the average MSE are respectively improved by 69% and 55.02%, and compared with the EMD-ELM method, the EEMD-approximate entropy-ELM prediction method has the advantages that the average MAPE and the average MSE are respectively improved by 65.6% and 44.1%. Among the three models, the model provided by the embodiment has the highest prediction precision and better overall prediction effect.
TABLE 2 error index of three models
Figure BDA0002277510900000111
Therefore, the short-term load prediction model of the extreme learning machine based on the EEMD-approximate entropy proposed by the embodiment utilizes the EEMD to decompose the load, and effectively solves the modal aliasing problem existing in the EMD; then, similar components are added by calculating the approximate entropy of each component to obtain a new component, so that the difficulty of load prediction is reduced; and finally, respectively predicting the 3 new components by adopting ELM, and superposing the results. The experimental simulation result shows that the method of the embodiment has better prediction effect and is an effective load prediction method.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (6)

1. A method for predicting a power load, comprising:
decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components;
calculating the approximate entropy of each modal component, and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
respectively carrying out load prediction on each component sequence by using a load prediction model based on an extreme learning machine to obtain a load prediction result corresponding to each component sequence;
and superposing the load prediction results corresponding to each component sequence to obtain a load prediction value corresponding to the original load sequence.
2. The method according to claim 1, wherein the calculating the approximate entropies of the modal components and the superimposing the approximate entropies of each modal component to obtain a component sequence corresponding to each modal component comprises:
performing m-dimensional reconstruction on the one-dimensional time sequence of each modal component to obtain vectors V (j) and V (q);
calculating the distance d [ V (j), V (q) ] of two vectors V (j), V (q);
according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure FDA0002277510890000011
Calculating the average autocorrelation degree of the vector V (j) according to the correlation integral;
calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j);
and overlapping the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
3. The power load prediction method according to claim 1, wherein the performing the load prediction on each component sequence by using the extreme learning machine-based load prediction model to obtain the load prediction result corresponding to each component sequence comprises:
using the sample composed of the component sequences as an input vector of the extreme learning machine-based load prediction model, and using the expected output value corresponding to each component sequence as a load prediction result corresponding to each component sequence, wherein the load prediction model of the extreme learning machine is as follows:
Figure FDA0002277510890000021
wherein the content of the first and second substances,
Figure FDA0002277510890000022
g (x) is the activation function,
Figure FDA0002277510890000023
for the weights of the input layer to the hidden layer,
Figure FDA0002277510890000024
for the deviation of the i-th implicit node,
Figure FDA0002277510890000025
the weights for the hidden layer to the output layer,
Figure FDA0002277510890000026
for the sequence of components of the input the,
Figure FDA0002277510890000027
n is the number of samples for the desired output value.
4. An electrical load prediction system, comprising: the system comprises a decomposition module, a component sequence calculation module, a load prediction module and a load superposition module;
the decomposition module is used for decomposing the original load sequence by adopting a lumped empirical mode decomposition method to obtain a plurality of modal components;
the component sequence calculation module is used for calculating the approximate entropy of each modal component and superposing the approximate entropy of each modal component to obtain a component sequence corresponding to each modal component;
the load prediction module is used for respectively predicting the load of each component sequence by using a load prediction model based on the extreme learning machine to obtain a load prediction result corresponding to each component sequence;
and the load superposition module is used for superposing the load prediction results corresponding to each component sequence to obtain the load prediction value corresponding to the original load sequence.
5. The power load prediction system according to claim 4, wherein the component sequence calculation module includes a reconstruction unit, an inter-vector distance calculation unit, a correlation integral calculation unit, an average degree of autocorrelation calculation unit, an approximate entropy calculation unit, and a superposition unit;
the reconstruction unit is used for performing m-dimensional reconstruction on the one-dimensional time sequence of each modal component to obtain vectors V (j) and V (q);
the inter-vector distance calculation unit is used for calculating the distance d [ V (j), V (q) ] of two vectors V (j), V (q);
the correlation integral calculation unit is used for calculating the correlation integral according to the distance d [ V (j), V (q)]Calculating a correlation integral
Figure FDA0002277510890000028
The average autocorrelation degree calculating unit is used for calculating the average autocorrelation degree of the vector V (j) according to the correlation integral;
the approximate entropy calculation unit is used for calculating the approximate entropy of the modal component according to the average autocorrelation degree of the vector V (j);
the superposition unit is used for superposing the approximate entropies of the modal components to obtain a component sequence corresponding to each modal component.
6. The power load prediction system of claim 4, wherein the load prediction module comprises a sample construction unit and a prediction unit;
the sample construction unit is used for constructing samples according to the component sequences;
the prediction unit is used for taking the samples as input vectors of the extreme learning machine-based load prediction model, and taking expected output values corresponding to each component sequence as load prediction results corresponding to each component sequence, wherein the load prediction model of the extreme learning machine is as follows:
Figure FDA0002277510890000031
wherein the content of the first and second substances,
Figure FDA0002277510890000032
g (x) is the activation function,
Figure FDA0002277510890000033
for the weights of the input layer to the hidden layer,
Figure FDA0002277510890000034
for the deviation of the i-th implicit node,
Figure FDA0002277510890000035
the weights for the hidden layer to the output layer,
Figure FDA0002277510890000036
for the sequence of components of the input the,
Figure FDA0002277510890000037
n is the number of samples for the desired output value.
CN201911128175.XA 2019-11-18 2019-11-18 Power load prediction method and system Pending CN111126659A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911128175.XA CN111126659A (en) 2019-11-18 2019-11-18 Power load prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911128175.XA CN111126659A (en) 2019-11-18 2019-11-18 Power load prediction method and system

Publications (1)

Publication Number Publication Date
CN111126659A true CN111126659A (en) 2020-05-08

Family

ID=70495711

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911128175.XA Pending CN111126659A (en) 2019-11-18 2019-11-18 Power load prediction method and system

Country Status (1)

Country Link
CN (1) CN111126659A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113159361A (en) * 2020-12-03 2021-07-23 安徽大学 Short-term load prediction method and system based on VDM and Stacking model fusion

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104299043A (en) * 2014-06-13 2015-01-21 国家电网公司 Ultra-short-term load prediction method of extreme learning machine
CN105404939A (en) * 2015-12-04 2016-03-16 河南许继仪表有限公司 Short-term power load prediction method
CN107392364A (en) * 2017-07-12 2017-11-24 河海大学 The short-term load forecasting method of variation mode decomposition and depth belief network
CN109146183A (en) * 2018-08-24 2019-01-04 广东工业大学 Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm
CN109711755A (en) * 2019-01-23 2019-05-03 华南理工大学 Short-term power load prediction model establishment method based on EMD-VMD-PSO-LSSVM
CN109726865A (en) * 2018-12-27 2019-05-07 国网江苏省电力有限公司电力科学研究院 Customer charge probability density prediction technique, device and storage medium based on EMD-QRF
CN110009125A (en) * 2019-01-23 2019-07-12 华南理工大学 Short-term power load prediction model establishment method based on EMD-VMD-PSO-BPNN

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104299043A (en) * 2014-06-13 2015-01-21 国家电网公司 Ultra-short-term load prediction method of extreme learning machine
CN105404939A (en) * 2015-12-04 2016-03-16 河南许继仪表有限公司 Short-term power load prediction method
CN107392364A (en) * 2017-07-12 2017-11-24 河海大学 The short-term load forecasting method of variation mode decomposition and depth belief network
CN109146183A (en) * 2018-08-24 2019-01-04 广东工业大学 Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm
CN109726865A (en) * 2018-12-27 2019-05-07 国网江苏省电力有限公司电力科学研究院 Customer charge probability density prediction technique, device and storage medium based on EMD-QRF
CN109711755A (en) * 2019-01-23 2019-05-03 华南理工大学 Short-term power load prediction model establishment method based on EMD-VMD-PSO-LSSVM
CN110009125A (en) * 2019-01-23 2019-07-12 华南理工大学 Short-term power load prediction model establishment method based on EMD-VMD-PSO-BPNN

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
郭瑞等: "一种结合互补集合经验模态分解和小波核极限学习机的短期电力负荷预测模型", 《计算机应用与软件》 *
陈艳平等: "基于EEMD-样本熵和Elman神经网络的短期电力负荷预测", 《电力系统及其自动化学报》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113159361A (en) * 2020-12-03 2021-07-23 安徽大学 Short-term load prediction method and system based on VDM and Stacking model fusion

Similar Documents

Publication Publication Date Title
CN110309603B (en) Short-term wind speed prediction method and system based on wind speed characteristics
CN111860982A (en) Wind power plant short-term wind power prediction method based on VMD-FCM-GRU
CN109088407B (en) Power distribution network state estimation method based on deep belief network pseudo-measurement modeling
CN110991721A (en) Short-term wind speed prediction method based on improved empirical mode decomposition and support vector machine
CN114707712A (en) Method for predicting requirement of generator set spare parts
CN114091766B (en) CEEMDAN-LSTM-based space load prediction method
CN116665483A (en) Novel method for predicting residual parking space
CN117200208B (en) User-level short-term load prediction method and system based on multi-scale component feature learning
CN111126659A (en) Power load prediction method and system
Dong et al. Short-term load forecasting based on EEMD-approximate entropy and ELM
CN117633446A (en) Wind speed prediction method and system based on CEEMD-SE-EWT double decomposition and ARIMA combined prediction
CN117592593A (en) Short-term power load prediction method based on improved quadratic modal decomposition and WOA optimization BILSTM-intent
CN117077327A (en) Bearing life prediction method and system based on digital twin
CN116995670A (en) Photovoltaic power ultra-short-term prediction method based on multi-mode decomposition and multi-branch input
CN112132328A (en) Photovoltaic output power ultra-short-term local emotion reconstruction neural network prediction method
CN115326397B (en) Method and related device for establishing crankshaft bearing wear degree prediction model and prediction method
CN115809725A (en) Multi-factor short-term electric quantity prediction method and device
CN116316600A (en) Wind power day-ahead prediction self-adaptive filtering method based on seasonal typing
CN114638421A (en) Method for predicting requirement of generator set spare parts
Guo et al. Information fusion and XGBoost algorithm used for bearing remaining useful life prediction
Zhu et al. Prediction of air quality index based on wavelet transform combination model
Yan et al. AQI prediction based on CEEMD-WOA-Elman neural network
Chen et al. A Hybrid Model for Time Series Prediction Using Adaptive Variational Mode Decomposition
CN117494573B (en) Wind speed prediction method and system and electronic equipment
Zhang et al. An Artificial Intelligent Method of Power Load Forecasting in Short-term

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200508