CN111931973A - Method for improving short-time load prediction precision of cable line - Google Patents
Method for improving short-time load prediction precision of cable line Download PDFInfo
- Publication number
- CN111931973A CN111931973A CN202010550278.1A CN202010550278A CN111931973A CN 111931973 A CN111931973 A CN 111931973A CN 202010550278 A CN202010550278 A CN 202010550278A CN 111931973 A CN111931973 A CN 111931973A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- output
- model
- training
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000003062 neural network model Methods 0.000 claims abstract description 82
- 239000002245 particle Substances 0.000 claims abstract description 52
- 239000013598 vector Substances 0.000 claims abstract description 34
- 238000012360 testing method Methods 0.000 claims abstract description 32
- 238000012706 support-vector machine Methods 0.000 claims abstract description 29
- 238000005457 optimization Methods 0.000 claims abstract description 12
- 238000007689 inspection Methods 0.000 claims abstract description 4
- 210000002569 neuron Anatomy 0.000 claims description 29
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 230000004913 activation Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 230000014509 gene expression Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
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
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A method for improving the short-time load prediction precision of a cable line comprises the following steps: s1, acquiring historical data of short-time load of a cable line to form a training sample and an inspection sample; s2, respectively constructing a BP neural network model, a Chebyshev neural network model and a support vector machine model by using training samples, and training each model; s3, respectively inputting the test samples into the trained BP neural network model, the Chebyshev neural network model and the support vector machine model to obtain the output of each model; s4, performing weighted combination by utilizing the output of each model, and combining a particle swarm optimization algorithm to obtain the corresponding weight of the output of each model; s5, inputting input vectors corresponding to moments to be predicted into the trained BP neural network model, the Chebyshev neural network model and the support vector machine model to obtain corresponding outputs of the models; and combining the weights obtained in the step S4 to obtain a final short-time load predicted value. The method provided by the invention can further improve the accuracy of the short-time load prediction of the cable line.
Description
Technical Field
The invention relates to the technical field of short-time load prediction, in particular to a method for improving the short-time load prediction precision of a cable line.
Background
Load forecasting may be divided into long-term, medium-term, and short-term load forecasting by time of forecasting. The accurate prediction of the short-term load is beneficial to improving the operation benefit of the power operation main body.
At present, the method for predicting the short-term load of the power system mainly comprises a statistical technique, an expert system method and a neural network method. Short-term load models used in statistical techniques can be generally classified into time system models and regression models. The time system model cannot make full use of climate information and other factors that have a large impact on load performance, and predictions are inaccurate and unstable. The regression model needs to know the functional relationship between the load and the meteorological variable in advance, and has large calculation amount, and can not process the non-equilibrium transient relationship between the meteorological variable and the load. The expert system method utilizes the experience knowledge and the inference rule of experts, improves the load prediction precision of holidays or major activity days, but has great difficulty in accurately converting the expert knowledge, the experience and the like into a series of rules.
Recent studies have shown that, compared with statistical techniques and expert system methods, as described in chinese patent CN 107301478A, the prediction of the short-term load of the power system by using the neural network method can achieve higher accuracy, but there is a limit to the prediction of the short-term load of the power system by using the neural network method alone.
Disclosure of Invention
The invention provides a method for improving the short-term load prediction precision of a cable line, aiming at overcoming the defect that the short-term load prediction of a power system by using a neural network method in the prior art still has a certain limit. The method performs weighted combination on the predicted values of the BP neural network, the Chebyshev neural network and the support vector machine, and determines corresponding weights by using a particle swarm optimization algorithm so as to achieve the purpose of improving the short-time load prediction precision of the cable line.
In order to solve the technical problems, the invention adopts the technical scheme that: a method for improving the load prediction precision of a cable short-time circuit comprises the following steps:
s1, acquiring historical data of short-time load of a cable line to form a training sample and an inspection sample;
s2, respectively constructing a BP neural network model, a Chebyshev neural network model and a support vector machine model by using training samples, and training the models to obtain the trained models;
s3, respectively inputting the test samples into the trained BP neural network model, the Chebyshev neural network model and the support vector machine model to obtain the output of each model;
s4, performing weighted combination by utilizing the output of each model, and combining a particle swarm optimization algorithm to obtain the corresponding weight of the output of each model;
s5, inputting the input vector corresponding to the moment to be predicted into the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain the corresponding output of each model; and combining the weights obtained in the step S4 to obtain a final short-time load predicted value of the time to be predicted.
Further, in step S1, the cable short-time load history data sequence is recorded as It(T ═ 1, 2.., T), then the training and test samples are denoted (I)t-6,It-5,It-4,It-3,It-2,It-1,It) Wherein (I)t-6,It-5,It-4,It-3,It-2,It-1) As an input vector, ItIs the output. Although the expressions of the training sample and the test sample are the same, the values of t corresponding to the training sample and the test sample are different.
Further, in the step S2, constructing the BP neural network model by using the training samples specifically includes the following steps:
s211, constructing a BP neural network model by using the training samples, wherein the BP neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, X and O1Representing input and output variables, W, of the network, respectively1 TThe weight matrices of the input layer and the first hidden layer,the weight matrices for the first hidden layer and the second hidden layer,is a weight matrix of the nth hidden layer and the output layer, f is an activation function of the hidden layer, purelin is an activation function of the output layer, b1And b2A threshold vector for the hidden layer, bn+1Is the threshold vector of the output layer;
s212, determining the number of neurons in an input layer of the BP neural network model according to an input vector, determining the number of neurons in an output layer according to an output vector, and determining the number of hidden layers and the number of neurons in each hidden layer according to a trial and error method, namely training the network, then checking the generalization capability of the network, and further determining the number of hidden layers and the number of neurons in each hidden layer;
s213, training the BP neural network model by using the training sample to obtain the trained BP neural network model.
Further, in the step S2, constructing the Chebyshev neural network model by using the training sample specifically includes the following steps:
s221, constructing a Chebyshev neural network model by using the training samples, wherein the Chebyshev neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, XiAnd OpRespectively representing the ith element of the input vector X and the pth element of the output vector O, wijIs an input layer and an output layerConnection weight, cpjConnecting weights for hidden and output layers, Tj(x) The order is a Chebyshev neural network orthogonal polynomial, J is the order of the Chebyshev neural network orthogonal polynomial, and I is the number of input layer neurons; wherein T isj(x) The expression of (a) is:
s222, determining the number of neurons in an input layer of a Chebyshev neural network model according to an input vector, determining the number of neurons in an output layer according to an output vector, and determining the number of neurons in a hidden layer according to a trial and error method, namely training a network, then testing the generalization capability of the network, and further determining the number of neurons in the hidden layer;
and S223, training the Chebyshev neural network model by using the training sample to obtain the trained Chebyshev neural network model.
Further, in step S3, the test sample is input to the trained BP neural network model, and the output of the BP neural network model is I'1(t)(t=1,2,..,T1),T1To test the number of samples.
Further, in step S3, the test sample is input into the trained Chebyshev neural network model, and the output of the Chebyshev neural network model is I'2(t)(t=1,2,..,T1),T1To check the number of samples.
Further, in step S3, the test sample is input into the trained support vector machine model, and the output of the support vector machine model is I'3(t)(t=1,2,..,T1),T1To test the number of samples.
Further, the step S4 specifically includes the following steps:
s41, setting the actual value of the test sample as I' (t), and carrying out weighted combination on the outputs of the models:
I″(t)=ρt1×I′1(t)+ρt2×I′2(t)+ρt3×I′3(t)
in the formula, ρt1、ρt2、ρt3Corresponding weights output by the BP neural network model, the Chebyshev neural network model and the support vector machine model at the time t respectively, and rhot1+ρt2+ρt3=1;
S42, weighting rho by using particle swarm optimization algorithmt1、ρt2、ρt3And (6) optimizing.
Further, the step S42 specifically includes the following steps:
s421, initializing particle swarm, namely assigning an initial value to each group of weights; setting the size of the particle swarm and the maximum iteration times, defining a fitness function of the particle swarm as an error between an output combination value and an actual value, and calculating the fitness of each particle as shown in the following formula;
s422, determining a local extreme value and a global extreme value; firstly, comparing the current fitness of each particle with the previous optimal fitness, and taking the current local extreme value of each particle as the smaller value of the current local extreme value and the previous optimal fitness; when determining the global extreme value, taking the minimum one of the current fitness of all the particles;
s423, carrying out iterative computation, wherein in each iteration, the speed and the position of the particles are respectively updated according to the following formula by using the local extreme value and the global extreme value obtained in the step S422;
the velocity update formula for the particles is as follows:
vij(k+1)=wvij(k)+c1r1[Qij(k)-xij(k)] +c2r2[Qgj(k)-xij(k)]
the particle position update formula is as follows:
xij(k+1)=xij(k)+vij(k+1)
in the formula, vijAnd xijRespectively the speed and the position of the jth dimension space of the ith particle; qijFor local extrema, Q, of the jth dimension of the ith particlegjA global extreme value of a j-dimensional space of all the particles is obtained; w is an inertial weight coefficient, c1And c2Is an acceleration constant, r1And r2Is a random number, and the value interval is [ 01 ]];
S424, when the algorithm meets the error precision or reaches the maximum iteration times, exiting the particle swarm optimization algorithm;
s425, outputting optimized weight rho't1、ρ′t2、ρ′t3。
Further, the step S5 specifically includes the following steps:
s51, inputting the corresponding input vector (I) of the moment to be predictedT-5,IT-4,IT-3,IT-2,IT-1,IT) Inputting the model into the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain corresponding outputs of the models, which are respectively marked as I1(T+1)、I2(T+1)、I3(T+1);
S52. according to I1(T+1)、I2(T+1)、I3(T +1), in combination of ρ't1、ρ′t2、ρ′t3Obtaining a final load predicted value I (T + 1):
I(T+1)=I1(T+1)×ρ′t1+I2(T+1)×ρ′t2+I3(T+1)×ρ′t3。
compared with the prior art, the invention has the beneficial effects that:
the method for improving the cable short-time load prediction precision provided by the invention is characterized in that weighting combination is carried out on the predicted values of the BP neural network, the Chebyshev neural network and the support vector machine, corresponding weights are determined by using a particle swarm optimization algorithm, the short-time load predicted value is finally obtained, and the cable short-time load prediction precision is further improved.
Drawings
Fig. 1 is a schematic flow chart of a method for improving the short-time load prediction accuracy of a cable line according to the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent; for a better understanding of the present embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent actual product dimensions; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted. The positional relationships depicted in the drawings are for illustrative purposes only and are not to be construed as limiting the present patent.
As shown in fig. 1, a method for improving the accuracy of cable line short-time load prediction includes the following steps:
s1, acquiring historical data of short-time load of a cable line to form a training sample and an inspection sample; recording the cable line short-time load historical data sequence as It(T ═ 1, 2.., T), then the training and test samples are denoted (I)t-6,It-5, It-4,It-3,It-2,It-1,It) Wherein (I)t-6,It-5,It-4,It-3,It-2,It-1) As an input vector, ItIs an output; wherein the number of training samples is 800, and the number of testing samples is 200.
And S2, respectively constructing a BP neural network model, a Chebyshev neural network model and a support vector machine model by using the training samples, and training the models to obtain the trained models. The method specifically comprises the following steps:
s211, constructing a BP neural network model by using the training samples, wherein the BP neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, X and O1Representing input and output variables, W, of the network, respectively1 TWeights for the input layer and the first hidden layerThe matrix is a matrix of a plurality of matrices,the weight matrices for the first hidden layer and the second hidden layer,is a weight matrix of the nth hidden layer and the output layer, f is an activation function of the hidden layer, purelin is an activation function of the output layer, b1And b2A threshold vector for the hidden layer, bn+1Is the threshold vector of the output layer;
s212, determining the number of neurons in an input layer of the BP neural network model according to an input vector, wherein the number is 6 in the embodiment; the number of neurons in the output layer is determined according to the output vector, which is 1 in this embodiment; the number of hidden layers and the number of neurons of each hidden layer are determined according to a trial and error method, namely, training the network, and then checking the generalization ability of the network to further determine the number of the hidden layers and the number of neurons of each hidden layer; through repeated trial and error, the number of hidden layers is determined to be 2, and the number of neurons in each hidden layer is 12 and 14 respectively.
S213, training the BP neural network model by using the training sample to obtain the trained BP neural network model.
S221, constructing a Chebyshev neural network model by using the training samples, wherein the Chebyshev neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, XiAnd OpRespectively representing the ith element of the input vector X and the pth element of the output vector O, wijConnecting weights for input and output layers, cpjConnecting weights for hidden and output layers, Tj(x) The order is a Chebyshev neural network orthogonal polynomial, J is the order of the Chebyshev neural network orthogonal polynomial, and I is the number of input layer neurons; wherein T isj(x) The expression of (a) is:
s222, determining the number of neurons of an input layer of the Chebyshev neural network model according to an input vector, wherein the number is 6 in the embodiment; the number of neurons in the output layer is determined according to the output vector, which is 1 in this embodiment; determining the number of neurons of the hidden layer according to a trial and error method, namely training the network, and then checking the generalization ability of the network to determine the number of neurons of the hidden layer; through repeated trial and error, the number of hidden layer neurons is 12.
And S223, training the Chebyshev neural network model by using the training sample to obtain the trained Chebyshev neural network model.
S231, constructing a support vector machine by using the training samples, and training the support vector machine to obtain the trained support vector machine.
And S3, respectively inputting the test samples into the trained BP neural network model, the Chebyshev neural network model and the support vector machine model to obtain the output of each model. The method specifically comprises the following steps:
s31, inputting the test sample into the trained BP neural network model to obtain the output I of the BP neural network model1′(t)(t=1,2,..,T1),T1To test the number of samples, 200 in this example.
S32, inputting the test sample into the trained Chebyshev neural network model to obtain an output I of the Chebyshev neural network model2′(t)(t=1,2,..,T1),T1To test the number of samples, 200 in this example.
S33, inputting the test sample into the trained support vector machine model to obtain I 'of the output of the support vector machine model'3(t)(t=1,2,..,T1),T1To test the number of samples, 200 in this example.
And S4, performing weighted combination by utilizing the output of each model, and combining a particle swarm optimization algorithm to obtain the corresponding weight of the output of each model. The method specifically comprises the following steps:
s41, setting the actual value of the test sample as I' (t), and carrying out weighted combination on the outputs of the models:
I″(t)=ρt1×I′1(t)+ρt2×I′2(t)+ρt3×I′3(t)
in the formula, ρt1、ρt2、ρt3Corresponding weights output by the BP neural network model, the Chebyshev neural network model and the support vector machine model at the time t respectively, and rhot1+ρt2+ρt3=1;
S42, weighting rho by using particle swarm optimization algorithmt1、ρt2、ρt3And (6) optimizing. The method specifically comprises the following steps:
s421, initializing particle swarm, namely assigning an initial value to each group of weights; setting the particle swarm size and the maximum iteration number, wherein in the example, the particle swarm size is 40, and the maximum iteration number is 500; the fitness function of the particle swarm is defined as an output combination value and an actual value error, and the fitness of each particle is calculated as shown in the following formula.
S422, determining a local extreme value and a global extreme value; firstly, comparing the current fitness of each particle with the previous optimal fitness, and taking the current local extreme value of each particle as the smaller value of the current local extreme value and the previous optimal fitness; when determining the global extreme value, taking the minimum one of the current fitness of all the particles;
s423, carrying out iterative computation, wherein in each iteration, the speed and the position of the particles are respectively updated according to the following formula by using the local extreme value and the global extreme value obtained in the step S422;
the velocity update formula for the particles is as follows:
vij(k+1)=wvij(k)+c1r1[Qij(k)-xij(k)] +c2r2[Qgj(k)-xij(k)]
the particle position update formula is as follows:
xij(k+1)=xij(k)+vij(k+1)
in the formula, vijAnd xijRespectively the speed and the position of the jth dimension space of the ith particle; qijFor local extrema, Q, of the jth dimension of the ith particlegjA global extreme value of a j-dimensional space of all the particles is obtained; w is an inertial weight coefficient, c1And c2Is an acceleration constant, r1And r2Is a random number, and the value interval is [ 01 ]];
S424, when the algorithm meets the error precision or reaches the maximum iteration times, exiting the particle swarm optimization algorithm;
s425, outputting optimized weight rho't1、ρ′t2、ρ′t3。
S5, inputting the input vector corresponding to the moment to be predicted into the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain the corresponding output of each model; and combining the weights obtained in the step S4 to obtain a final short-time load predicted value of the time to be predicted. The method specifically comprises the following steps:
s51, inputting the corresponding input vector (I) of the moment to be predictedT-5,IT-4,IT-3,IT-2,IT-1,IT) Inputting the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain corresponding outputs of the models, which are respectively marked as I1(T+1)、I2(T+1)、I3(T+1);
S52. according to I1(T+1)、I2(T+1)、I3(T +1), in combination of ρ't1、ρ′t2、ρ′t3Obtaining a final load predicted value I (T + 1):
I(T+1)=I1(T+1)×ρ′t1+I2(T+1)×ρ′t2+I3(T+1)×ρ′t3。
it should be understood that the above-described examples are merely illustrative for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. It will be apparent to those skilled in the art that other variations and modifications can be made on the basis of the above description. This need not be, nor should it be exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A method for improving the short-time load prediction precision of a cable line is characterized by comprising the following steps:
s1, acquiring historical data of short-time load of a cable line to form a training sample and an inspection sample;
s2, respectively constructing a BP neural network model, a Chebyshev neural network model and a support vector machine model by using training samples, and training the models to obtain the trained models;
s3, respectively inputting the test samples into the trained BP neural network model, the Chebyshev neural network model and the support vector machine model to obtain the output of each model;
s4, performing weighted combination by utilizing the output of each model, and combining a particle swarm optimization algorithm to obtain the corresponding weight of the output of each model;
s5, inputting the input vector corresponding to the moment to be predicted into the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain the corresponding output of each model; and combining the weights obtained in the step S4 to obtain a final short-time load predicted value of the time to be predicted.
2. The method for improving the cable short-time load prediction accuracy as claimed in claim 1, wherein in step S1, the cable short-time load history data sequence is recorded as It(T ═ 1, 2.., T), then the training and test samples are denoted (I)t-6,It-5,It-4,It-3,It-2,It-1,It) Wherein (I)t-6,It-5,It-4,It-3,It-2,It-1) As an input vector,ItIs the output.
3. The method for improving the cable line short-time load prediction accuracy according to claim 2, wherein in the step S2, the constructing the BP neural network model by using the training samples specifically includes the following steps:
s211, constructing a BP neural network model by using the training samples, wherein the BP neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, X and O1Representing input and output variables, W, of the network, respectively1 TThe weight matrices of the input layer and the first hidden layer,the weight matrices for the first hidden layer and the second hidden layer,is a weight matrix of the nth hidden layer and the output layer, f is an activation function of the hidden layer, purelin is an activation function of the output layer, b1And b2A threshold vector for the hidden layer, bn+1Is the threshold vector of the output layer;
s212, determining the number of neurons in an input layer of the BP neural network model according to an input vector, determining the number of neurons in an output layer according to an output vector, and determining the number of hidden layers and the number of neurons in each hidden layer according to a trial and error method, namely training the network, then checking the generalization capability of the network, and further determining the number of hidden layers and the number of neurons in each hidden layer;
s213, training the BP neural network model by using the training sample to obtain the trained BP neural network model.
4. The method for improving the cable line short-time load prediction accuracy as claimed in claim 3, wherein the step S2 of constructing the Chebyshev neural network model by using the training samples specifically includes the following steps:
s221, constructing a Chebyshev neural network model by using the training samples, wherein the Chebyshev neural network model comprises an input layer, a hidden layer and an output layer, and the output is as follows:
in the formula, XiAnd OpRespectively representing the ith element of the input vector X and the pth element of the output vector O, wijConnecting weights for input and output layers, cpjConnecting weights for hidden and output layers, Tj(x) The order is a Chebyshev neural network orthogonal polynomial, J is the order of the Chebyshev neural network orthogonal polynomial, and I is the number of neurons in an input layer; wherein T isj(x) The expression of (a) is:
s222, determining the number of neurons in an input layer of a Chebyshev neural network model according to an input vector, determining the number of neurons in an output layer according to an output vector, and determining the number of neurons in a hidden layer according to a trial and error method, namely training a network, then testing the generalization ability of the network, and further determining the number of neurons in the hidden layer;
and S223, training the Chebyshev neural network model by using the training sample to obtain the trained Chebyshev neural network model.
5. The method for improving the cable line short-time load prediction accuracy as claimed in claim 4, wherein in the step S3, the test sample is input into the trained BP neural network model, and the output of the BP neural network model is I'1(t)(t=1,2,..,T1),T1For testing the number of samples。
6. The method for improving the cable line short-time load prediction accuracy as claimed in claim 5, wherein in the step S3, the test sample is input into the trained Chebyshev neural network model, and the output of the Chebyshev neural network model is I'2(t)(t=1,2,..,T1),T1To test the number of samples.
7. The method for improving the cable line short-time load prediction accuracy as claimed in claim 6, wherein in the step S3, the test sample is input into the trained support vector machine model, and the output of the support vector machine model is I'3(t)(t=1,2,..,T1),T1To test the number of samples.
8. The method for improving the cable line short-time load prediction accuracy as claimed in claim 7, wherein the step S4 specifically comprises the following steps:
s41, setting the actual value of the test sample as I' (t), and carrying out weighted combination on the outputs of the models:
I″(t)=ρt1×I′1(t)+ρt2×I′2(t)+ρt3×I′3(t)
in the formula, ρt1、ρt2、ρt3Corresponding weights output by the BP neural network model, the Chebyshev neural network model and the support vector machine model at the time t respectively, and rhot1+ρt2+ρt3=1;
S42, weighting rho by using particle swarm optimization algorithmt1、ρt2、ρt3And (6) optimizing.
9. The method for improving the cable line short-time load prediction accuracy as claimed in claim 8, wherein the step S42 specifically comprises the following steps:
s421, initializing particle swarm, namely assigning an initial value to each group of weights; setting the size of the particle swarm and the maximum iteration times, defining a fitness function of the particle swarm as an output combination value and an actual value error, and calculating the fitness of each particle as shown in the following formula;
s422, determining a local extreme value and a global extreme value; firstly, comparing the current fitness of each particle with the previous optimal fitness, and taking the current local extreme value of each particle as the smaller value of the current local extreme value and the previous optimal fitness; when the global extreme value is determined, the minimum one of the current fitness of all the particles is selected;
s423, carrying out iterative computation, wherein in each iteration, the speed and the position of the particles are respectively updated according to the following formula by using the local extreme value and the global extreme value obtained in the step S422;
the velocity update formula for the particles is as follows:
vij(k+1)=wvij(k)+c1r1[Qij(k)-xij(k)]+c2r2[Qgj(k)-xij(k)]
the particle position update formula is as follows:
xij(k+1)=xij(k)+vij(k+1)
in the formula, vijAnd xijRespectively the speed and the position of the jth dimension space of the ith particle; qijFor local extrema, Q, of the jth dimension of the ith particlegjA global extreme value of a j-dimensional space of all the particles is obtained; w is an inertial weight coefficient, c1And c2Is an acceleration constant, r1And r2Is a random number, and the value interval is [ 01 ]];
S424, when the algorithm meets the error precision or reaches the maximum iteration times, exiting the particle swarm optimization algorithm;
s425, outputting optimized weight rho't1、ρ′t2、ρ′t3。
10. The method for improving the cable line short-time load prediction accuracy as claimed in claim 9, wherein the step S5 specifically comprises the following steps:
s51, inputting the corresponding input vector (I) of the moment to be predictedT-5,IT-4,IT-3,IT-2,IT-1,IT) Inputting the model into the BP neural network model, the Chebyshev neural network model and the support vector machine model trained in the step S2 to obtain corresponding outputs of the models, which are respectively marked as I1(T+1)、I2(T+1)、I3(T+1);
S52. according to I1(T+1)、I2(T+1)、I3(T +1), in combination of ρ't1、ρ′t2、ρ′t3And obtaining a final load predicted value I (T + 1):
I(T+1)=I1(T+1)×ρ′t1+I2(T+1)×ρ′t2+I3(T+1)×ρ′t3。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010550278.1A CN111931973A (en) | 2020-06-16 | 2020-06-16 | Method for improving short-time load prediction precision of cable line |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010550278.1A CN111931973A (en) | 2020-06-16 | 2020-06-16 | Method for improving short-time load prediction precision of cable line |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111931973A true CN111931973A (en) | 2020-11-13 |
Family
ID=73317596
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010550278.1A Pending CN111931973A (en) | 2020-06-16 | 2020-06-16 | Method for improving short-time load prediction precision of cable line |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111931973A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112598201A (en) * | 2021-03-01 | 2021-04-02 | 湖北鑫英泰系统技术股份有限公司 | Power grid load estimation method and device based on distributed optical fiber temperature measurement |
CN112862168A (en) * | 2021-01-28 | 2021-05-28 | 中山大学 | Neural network multi-model combination-based population density prediction method and system |
CN113762629A (en) * | 2021-09-10 | 2021-12-07 | 南京机电职业技术学院 | Athlete optimal load strength prediction method based on improved AE |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106548025A (en) * | 2016-10-28 | 2017-03-29 | 广东电网有限责任公司珠海供电局 | A kind of the air gap discharge voltage Forecasting Methodology and device |
CN107301478A (en) * | 2017-06-26 | 2017-10-27 | 广东电网有限责任公司珠海供电局 | A kind of cable run short-term load forecasting method |
CN107590562A (en) * | 2017-09-05 | 2018-01-16 | 西安交通大学 | A kind of Short-Term Load Forecasting of Electric Power System based on changeable weight combination predicted method |
CN107766995A (en) * | 2017-12-08 | 2018-03-06 | 国网山东省电力公司枣庄供电公司 | Power-system short-term load forecasting method based on depth recurrent neural network |
-
2020
- 2020-06-16 CN CN202010550278.1A patent/CN111931973A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106548025A (en) * | 2016-10-28 | 2017-03-29 | 广东电网有限责任公司珠海供电局 | A kind of the air gap discharge voltage Forecasting Methodology and device |
CN107301478A (en) * | 2017-06-26 | 2017-10-27 | 广东电网有限责任公司珠海供电局 | A kind of cable run short-term load forecasting method |
CN107590562A (en) * | 2017-09-05 | 2018-01-16 | 西安交通大学 | A kind of Short-Term Load Forecasting of Electric Power System based on changeable weight combination predicted method |
CN107766995A (en) * | 2017-12-08 | 2018-03-06 | 国网山东省电力公司枣庄供电公司 | Power-system short-term load forecasting method based on depth recurrent neural network |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112862168A (en) * | 2021-01-28 | 2021-05-28 | 中山大学 | Neural network multi-model combination-based population density prediction method and system |
CN112598201A (en) * | 2021-03-01 | 2021-04-02 | 湖北鑫英泰系统技术股份有限公司 | Power grid load estimation method and device based on distributed optical fiber temperature measurement |
CN112598201B (en) * | 2021-03-01 | 2021-05-18 | 湖北鑫英泰系统技术股份有限公司 | Power grid load estimation method and device based on distributed optical fiber temperature measurement |
CN113762629A (en) * | 2021-09-10 | 2021-12-07 | 南京机电职业技术学院 | Athlete optimal load strength prediction method based on improved AE |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111970163B (en) | Network flow prediction method of LSTM model based on attention mechanism | |
CN107622303B (en) | Method for neural network and device for performing the method | |
US20190034784A1 (en) | Fixed-point training method for deep neural networks based on dynamic fixed-point conversion scheme | |
CN111931973A (en) | Method for improving short-time load prediction precision of cable line | |
TW202141363A (en) | Adaptive quantization for execution of machine learning models | |
US20230351180A1 (en) | Quantization recognition training method of neural network that supplements limitations of gradient-based learning by adding gradient-indipendent update | |
CN112257847A (en) | Method for predicting geomagnetic Kp index based on CNN and LSTM | |
CN115358305A (en) | Incremental learning robustness improving method based on iterative generation of boundary samples | |
CN114487890A (en) | Lithium battery health state estimation method for improving long-term and short-term memory neural network | |
CN113407820A (en) | Model training method, related system and storage medium | |
CN116303786B (en) | Block chain financial big data management system based on multidimensional data fusion algorithm | |
CN117273217A (en) | Space debris track forecasting method and system and electronic equipment | |
CN115062528A (en) | Prediction method for industrial process time sequence data | |
CN114897274A (en) | Method and system for improving time sequence prediction effect | |
Bazzi et al. | Comparative performance of several recent supervised learning algorithms | |
CN114372558A (en) | Residential electricity load prediction method, medium and equipment based on multi-model fusion | |
CN116324807A (en) | Neural architecture and hardware accelerator search | |
CN108960406B (en) | MEMS gyroscope random error prediction method based on BFO wavelet neural network | |
CN115516466A (en) | Hyper-parametric neural network integration | |
CN112529328A (en) | Product performance prediction method and system | |
Rajini et al. | Performance evaluation of neural networks for shape identification in image processing | |
CN111160419A (en) | Electronic transformer data classification prediction method and device based on deep learning | |
Li et al. | A self-adaptive global harmony search based extreme learning machine for classification problem | |
CN116956738B (en) | Mechanical structure reliability calculation method based on self-adaptive agent model | |
CN117852609A (en) | Method and device for preprocessing large language model |
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 |