CN115049136A - Transformer load prediction method - Google Patents
Transformer load prediction method Download PDFInfo
- Publication number
- CN115049136A CN115049136A CN202210718631.1A CN202210718631A CN115049136A CN 115049136 A CN115049136 A CN 115049136A CN 202210718631 A CN202210718631 A CN 202210718631A CN 115049136 A CN115049136 A CN 115049136A
- Authority
- CN
- China
- Prior art keywords
- samples
- load
- transformer
- data
- neural network
- 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 52
- 238000013528 artificial neural network Methods 0.000 claims abstract description 60
- 238000012549 training Methods 0.000 claims abstract description 37
- 230000002159 abnormal effect Effects 0.000 claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 7
- 238000003064 k means clustering Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 5
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000007621 cluster analysis Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 5
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 5
- 230000005284 excitation Effects 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 3
- 238000010187 selection method Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- 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
- 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
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Primary Health Care (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Power Engineering (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)
Abstract
The invention relates to a transformer load prediction method, and belongs to the technical field of power load prediction methods. The technical scheme of the invention is as follows: the method comprises the steps of taking a historical daily load curve of the transformer as an original sample, identifying and eliminating abnormal load curves by utilizing an improved clustering algorithm to obtain several types of typical daily load curve samples, taking approximate quantity samples from each type of the screened several types of samples to form a neural network training set, training the neural network and establishing an accurate prediction model of the daily load curve of the transformer. The invention has the beneficial effects that: the method eliminates the interference of abnormal data samples on the neural network, simultaneously realizes the screening and classification of available samples, reduces the scale of training samples, does not omit the class with smaller number of samples, and achieves the purpose that smaller training samples contain more useful characteristics, thereby improving the operation speed of the neural network, simultaneously improving the generalization performance and the prediction precision of a prediction model, and realizing more accurate prediction of the daily load of the transformer.
Description
Technical Field
The invention relates to a transformer load prediction method, and belongs to the technical field of power load prediction methods.
Background
The transformer daily load prediction is a key link of winding hot point temperature prediction and formulation of a daily mode, the traditional load prediction is generally based on a similar day or simplified load model, but the regional load has great relation with local economy, climate, industrial composition and electricity utilization habits, the electricity utilization laws in different regions are different, no universally applicable load prediction model exists, and the traditional method is insufficient in pertinence and prediction accuracy. With the rapid development of computer science and technology, the machine learning algorithm is used for learning and modeling historical load, weather and other data, the nonlinear mapping relation between input and output quantities is established in a targeted manner, and the load prediction precision in various occasions is greatly improved. The commonly used intelligent prediction method mainly comprises the following steps: grey system theory, expert system method, artificial neural network method, optimal combination prediction method and the like. The artificial neural network is the most basic method and the application of the artificial neural network is the most extensive.
An Artificial Neural Network (ANN) is a multi-layer network structure composed of a plurality of Artificial neurons and used for simulating the learning and processing processes of human brain on information. The neurons of each layer of the ANN are connected with each other, different connecting weights exist among different neurons, and the information law is reliably approximated by learning various kinds of information. The ANN has the capabilities of memory, learning and automatic adjustment, and for a multilayer artificial neural network, only a proper network structure needs to be designed and enough samples are adopted for training, so that any complex nonlinear model and system can be approached theoretically.
Compared with the traditional prediction method, the ANN-based short-term load prediction has higher prediction precision, but depending on the quality of the training sample, the influence of the training sample on the neural network even exceeds the structure of the neural network, so that the accurate sample is provided and the prediction precision is very important to improve. For a large number of original samples, if all the original samples are selected, the training speed of the neural network is too slow; if the samples are randomly drawn, a small number of useful samples may be omitted, resulting in poor stability of the obtained prediction model. And due to the influence of load impact, noise and abnormal data in the original sample, the prediction error of the model obtained by training is larger due to no effective processing.
Disclosure of Invention
The invention aims to provide a transformer load prediction method, which takes a transformer historical daily load curve as an original sample, utilizes an improved clustering algorithm to identify and eliminate abnormal load curves in the original sample, reduces the interference of the abnormal load curves on a neural network, clustering and screening to obtain several types of typical daily load curve samples, taking approximate quantity samples from each type of the screened several types of samples to form a neural network training set, can greatly reduce the scale of the training samples, simultaneously ensure that the classes with smaller sample numbers are not missed, realize the purpose of containing more useful features with smaller training sample size, thereby greatly improving the operation speed of the neural network, improving the generalization performance and the prediction precision of the prediction model, the method achieves the purposes of establishing an accurate transformer load prediction model and accurately predicting the daily load of the transformer, and effectively solves the problems in the background technology.
The technical scheme of the invention is as follows: a transformer load prediction method comprises the following steps:
a. determining the number of nodes of an input layer and an output layer by taking the integral point load value of a transformer in the day before the forecast, the highest temperature of the forecast day, the lowest temperature of the forecast day and the weather and type of the forecast day as input data and the integral point load value of the forecast day as output data, and constructing an ANN model;
b. predicting weather and type of the non-numerical quantity on the current day, and quantizing the non-numerical quantity into data in the interval of [0,1 ];
c. smoothing the instantaneous impact data in the historical load data by adopting wavelet threshold denoising;
d. carrying out normalization processing on input and output samples of the artificial neural network;
e. screening samples by adopting clustering analysis, eliminating abnormal load data, and dividing the samples into a plurality of typical categories;
f. selecting samples with similar quantity as training samples of the ANN in each class according to the quantity of the samples in each class in the typical classification;
g. carrying out dimension reduction processing on load sample data;
h. training an ANN load prediction model, adjusting the number of hidden layers to count the prediction effect, determining the optimal number of the hidden layers of the neural network, and training to obtain a daily load prediction model of the transformer;
i. and the input of the transformer load prediction model is updated on line in real time, and the load at the whole point in the future 24 hours is predicted on line according to the historical load in the previous 24 hours and the weather and air temperature forecast in the prediction period.
In the step a, an ANN model is constructed, wherein the ANN model comprises 28 input nodes in total including 24 integral point load values of 24 points on the day before the forecast day, the highest temperature of the forecast day, the lowest temperature of the forecast day, the weather and the type of the forecast day, and comprises 24 output nodes in total including 24 integral point load values of 24 points on the forecast day.
In the step e, classifying the samples based on an effective index k-means clustering method, introducing a judging process for the number of samples in each classification, judging that the samples are abnormal when the number of the samples is less than a set value, classifying the samples again after removing the samples, and performing an effective index calculation formula according to an equation (1) when the number of the clusters is k:
wherein S is j And C j Respectively a jth classified curve and a corresponding clustering center; c k1 And C k2 Two different clustering centers when the classification number is k; n is a radical of s Is the number of data in the data set.
In the step f, from various types of typical samples obtained by cluster analysis, an approximate number of samples are randomly selected from each type and added into a neural network training set, so that all the typical types of samples are ensured to be included.
And g, in order to reduce the number of nodes of the ANN and improve the calculation speed, reducing the dimension of load sample data to obtain 24-point integral point load data.
The invention has the beneficial effects that: the method comprises the steps of taking a historical daily load curve of the transformer as an original sample, identifying and eliminating the abnormal load curve by using an improved clustering algorithm, clustering and screening several types of typical daily load curve samples while reducing the interference of the abnormal curve on a neural network, and taking approximate quantity samples from each type of the screened several types of samples to form a neural network training set, so that the scale of the training samples can be greatly reduced, the types with smaller number samples are not omitted, and more useful characteristics are included in the smaller training sample scale, so that the operation speed of the neural network is greatly improved, the generalization performance and the prediction precision of a prediction model are improved, and the aims of establishing an accurate transformer load prediction model and accurately predicting the daily load of the transformer are fulfilled.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a wavelet threshold denoising result graph of the load curve of the present invention;
FIG. 3 is a flow chart of the improved k-means clustering algorithm of the present invention;
FIG. 4 is a graph of the improved k-means load clustering results of the present invention;
FIG. 5 is a diagram of a transformer neural network load prediction model architecture in accordance with the present invention;
FIG. 6 is a graph comparing the load prediction effect of the present invention with other methods.
Detailed Description
The technical solution of the present invention will be further described in detail with reference to the accompanying drawings and embodiments, which are preferred embodiments of the present invention. It is to be understood that the described embodiments are merely a subset of the embodiments of the invention, and not all embodiments; it should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A transformer load prediction method comprises the following steps:
a. determining the number of nodes of an input layer and an output layer by taking the integral point load value of a transformer in the day before the forecast, the highest temperature of the forecast day, the lowest temperature of the forecast day and the weather and type of the forecast day as input data and the integral point load value of the forecast day as output data, and constructing an ANN model;
b. predicting weather and type of the non-numerical quantity on the current day, and quantizing the non-numerical quantity into data in the interval of [0,1 ];
c. smoothing the instantaneous impact data in the historical load data by adopting wavelet threshold denoising;
d. carrying out normalization processing on input and output samples of the artificial neural network;
e. screening samples by adopting clustering analysis, eliminating abnormal load data, and dividing the samples into a plurality of typical categories;
f. selecting samples with similar quantity as training samples of the ANN in each class according to the quantity of the samples in each class in the typical classification;
g. carrying out dimension reduction processing on load sample data;
h. training an ANN load prediction model, adjusting the number of hidden layers to count the prediction effect, determining the optimal number of the hidden layers of the neural network, and training to obtain a daily load prediction model of the transformer;
i. and the input of the transformer load prediction model is updated on line in real time, and the load at the whole point in the future 24 hours is predicted on line according to the historical load in the previous 24 hours and the weather and air temperature forecast in the prediction period.
In the step a, an ANN model is constructed, wherein the ANN model comprises 28 input nodes in total including 24 integral point load values of 24 points on the day before the forecast day, the highest temperature of the forecast day, the lowest temperature of the forecast day, the weather and the type of the forecast day, and comprises 24 output nodes in total including 24 integral point load values of 24 points on the forecast day.
In the step e, classifying the samples based on an effective index k-means clustering method, introducing a judging process for the number of samples in each classification, judging that the samples are abnormal when the number of the samples is less than a set value, classifying the samples again after removing the samples, and performing an effective index calculation formula according to an equation (1) when the number of the clusters is k:
wherein S is j And C j Respectively a jth classified curve and a corresponding clustering center; c k1 And C k2 Two different clustering centers when the classification number is k; n is a radical of s Is the number of data in the data set.
In the step f, from various types of typical samples obtained by cluster analysis, an approximate number of samples are randomly selected from each type and added into a neural network training set, so that all the typical types of samples are ensured to be included.
And g, in order to reduce the number of nodes of the ANN and improve the calculation speed, reducing the dimension of load sample data to obtain 24-point integral point load data.
In practical application, the ANN model constructed in the step a comprises 28 input nodes in total including 24 output nodes in total including 24 point load values at 24 points on the predicted day.
In the step b, the holidays, the double-holidays and the working days are respectively quantized into 0, 0.5 and 1, and the weather is quantized according to the table 1;
TABLE 1 weather situation quantification Table
In the step c, denoising and smoothing the daily load curve by adopting a wavelet threshold, removing spike noise and impact data in the data, and setting the threshold according to a formula (1):
wherein M is the median of the absolute value of the wavelet decomposition coefficient of the first layerCounting; k G An adjustment coefficient being a standard variance of the gaussian noise; and N is a signal scale.
In the step d, normalization processing is carried out on the integral point load value of the transformer, the highest temperature of the forecast day and the lowest temperature of the forecast day according to an equation (2), and sample data are converted into a [0,1] interval;
wherein x is min And x max Minimum and maximum values of the input samples, respectively.
In the step e, classifying the samples based on an effective index k-means clustering method, introducing a judging process for the number of samples in each classification, judging that the samples are abnormal when the number of the samples is less than a set value, classifying the samples again after removing the samples, and performing an effective index calculation formula according to an equation (3) when the number of the clusters is k:
wherein S is j 、C j Respectively a jth classified curve and a corresponding clustering center; c k1 、C k2 Two different clustering centers when the classification number is k; n is a radical of s Is the number of data in the data set.
In the step f, of various types of samples obtained by clustering, an approximate number of samples are randomly selected from each type and added into an ANN training set, so that the samples of each typical type are ensured to be included;
in the step g, in order to reduce the number of nodes of the ANN and improve the calculation speed, reducing the dimension of load sample data to obtain 24-point integral point load data;
in the step h, determining the range of the ANN hidden layers according to the formula (4), and adjusting the number of the hidden layers within the range until the model prediction performance reaches the best, and determining the number of the ANN hidden layers;
wherein n is the number of input nodes; l is the number of output nodes; α is an integer between 1 and 10.
Example (b):
referring to fig. 1, the present invention is divided into the following steps:
the method comprises the steps that historical load, environment temperature, date and weather data of a transformer are used as input data, a load value of a transformer on a forecast day is used as output data, and 365 daily load curves of the transformer with 50MW capacity in a certain area in a certain year are organized to serve as original samples;
respectively quantizing holidays, double-holidays and working days into 0, 0.5 and 1, and quantizing the weather according to a method shown in the table 1 to ensure that non-numerical quantities are quantized into ANN available data;
TABLE 1 weather situation quantification Table
Smoothing all historical load data by adopting a wavelet threshold denoising method, filtering peaks and impact data in the data, and setting thresholds of each layer according to a formula (1):
wherein M is the median of the absolute value of the first-layer wavelet decomposition coefficient; k G An adjustment coefficient being a standard variance of the gaussian noise; n is a signal scale;
a result graph of the load data subjected to wavelet threshold denoising is shown in fig. 2, and after smoothing, peaks in the original data are filtered;
normalizing the integral point load value, the highest predicted day and the lowest temperature of the transformer according to the formula (2), and transforming the sample data into a [0,1] interval;
wherein x is min And x max Minimum and maximum values of the input samples, respectively.
Fifthly, as shown in fig. 3, an abnormal data judgment flow is added in the improved k-means clustering analysis flow, the classification with the number of samples less than a set value obtained by clustering is regarded as invalid classification, and the residual samples are clustered again after the invalid classification is deleted;
the formula for calculating the effective exponent when the number of clusters is k is performed according to equation (3):
wherein S is j 、C j Respectively a jth classified curve and a corresponding clustering center; c k1 、C k2 Two different clustering centers when the classification number is k; n is a radical of s Is the number of data in the data set.
After the load samples are clustered by the effective index k-means, the obtained load samples are classified as shown in fig. 4, 11 abnormal samples are screened out, and a total of 354 usable samples of 3 typical classes are obtained.
Reducing the dimension of the load sample data to obtain 24-point integral load data, and reducing the number of nodes of the ANN to improve the calculation speed;
seventhly, training a neural network by using samples obtained by three different methods, and testing the prediction performance of the neural network by using the same sample;
respectively selecting 10 test set samples from the three types of clustered load samples to form a test set sample;
the training set sample selection method comprises the following steps: the method comprises the following steps: the remaining 335 samples, except the test set samples, were included in the training set; the method 2 comprises the following steps: randomly drawing 60 out of the remaining 335 samples except the test set sample; the method 3 comprises the following steps: in addition to the test set samples, 20 samples were randomly drawn from each of three types (324 total) obtained in the present invention;
eighthly, as shown in FIG. 5, the ANN comprises 28 input nodes in total, wherein the load value of the 24-point integral point in the day before the forecast day, the highest temperature of the forecast day, the lowest temperature of the forecast day, the weather of the forecast day and the type of the forecast day are measured;
the ANN comprises 24 output nodes in total of 24 integral point load values at 24 points on a prediction day;
determining the range of the hidden layers according to the formula (4), and adjusting the number of the hidden layers in the range until the model prediction performance reaches the best, and determining the number of the hidden layers as the number of the ANN hidden layers;
weight w of input layer of neural network ij And the output layer weight v jk Is determined according to equation (5);
wherein x is i Inputting the ith input of the neural network input layer; h is j The output of the jth node of the hidden layer; y is k And y d_k The output of the kth node of the output layer and the expected output; f' is the derivative of the hidden layer excitation function to the hidden layer input quantity; f' is the derivative of the output layer excitation function to the input quantity of the output layer; η ∈ (0,1), which is the learning rate;
the hidden layer excitation function adopts a sigmoid function:
the output layer excitation function is:
F(x)＝x (7)
ninthly, performing performance evaluation on the prediction absolute error, the average relative error, the root mean square error and the correlation coefficient by the transformer load prediction model from 5 dimensions;
predicted value y p (a) With the true value y r (a) Is absoluteThe error is:
E p (a)＝y p (a)-y r (a) (8)
the average error of the multiple predicted values is:
in the formula, N c Is the number of error values.
The average relative error reflects the overall error of the multiple predictors:
root mean square error P RMSE Reflecting the degree of deviation of a plurality of predicted values from the true values, wherein the smaller the value of the predicted values is, the more stable the predicted result is, and the root mean square error is as follows:
and (3) a correlation coefficient R is introduced to evaluate the multi-day load overall prediction performance of the neural network, and the performance of the model is better as R approaches to 1.
And training the neural network by using training set samples obtained by three sample selection methods. Taking the load prediction results of three types of samples in the test set as an example, fig. 6 is a comparison of the load prediction effects of three days therein.
The comparison shows that the three types of loads in the method 1 have low prediction precision but have small difference; in the method 2, the prediction precision is low, particularly the prediction error of the few classes is large, because the classes with few samples account for less in the training samples during random selection and the training is insufficient; the average relative error and the mean square error of the method 3 are smaller than those of the former two methods, and the prediction errors of the two methods are closer, which shows that the load prediction precision and the stability of each point are improved after the sample optimization method is adopted.
In order to further compare and analyze the overall prediction performance of the model, the absolute error, the average error, the root mean square error and the correlation coefficient are counted by taking days as a unit, 30 prediction daily loads in a test set are predicted, and the performance indexes of the model built by the three methods are compared and shown in a table 2.
TABLE 2 comparison of overall prediction performance for different sample selection methods
Absolute error E of method 3 M Average relative error P M Mean square error P RMSE And the correlation coefficient R is obviously smaller, and the model training time T train The method is shorter, so that the comprehensive analysis of the above calculation results shows that after the training samples are subjected to wavelet threshold denoising smoothing and improved k-means clustering screening, the load prediction precision, stability and speed under different conditions are improved obviously, and the overall prediction performance is improved greatly.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes performed by the present specification and drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (5)
1. A transformer load prediction method is characterized by comprising the following steps:
a. determining the number of nodes of an input layer and an output layer by taking the integral point load value of a transformer in the day before the forecast, the highest temperature of the forecast day, the lowest temperature of the forecast day and the weather and type of the forecast day as input data and the integral point load value of the forecast day as output data, and constructing an ANN model;
b. predicting weather and type of the non-numerical quantity on the current day, and quantizing the non-numerical quantity into data in the interval of [0,1 ];
c. smoothing the instantaneous impact data in the historical load data by adopting wavelet threshold denoising;
d. carrying out normalization processing on input and output samples of the artificial neural network;
e. screening samples by adopting clustering analysis, eliminating abnormal load data, and dividing the samples into a plurality of typical categories;
f. selecting samples in each class as training samples of the ANN according to the number of samples in each class in the typical classification;
g. carrying out dimension reduction processing on load sample data;
h. training an ANN load prediction model, adjusting the number of hidden layers to count the prediction effect, determining the optimal number of the hidden layers of the neural network, and training to obtain a daily load prediction model of the transformer;
i. and the input of the transformer load prediction model is updated on line in real time, and the load at the whole point in the future 24 hours is predicted on line according to the historical load in the previous 24 hours and the weather and air temperature forecast in the prediction period.
2. The transformer load prediction method according to claim 1, wherein: in the step a, an ANN model is constructed, wherein the ANN model comprises 28 input nodes in total including 24 integral point load values of 24 points on the day before the forecast day, the highest temperature of the forecast day, the lowest temperature of the forecast day, the weather and the type of the forecast day, and comprises 24 output nodes in total including 24 integral point load values of 24 points on the forecast day.
3. The transformer load prediction method according to claim 1, wherein: in the step e, classifying the samples based on an effective index k-means clustering method, introducing a judging process for the number of samples in each classification, judging that the samples are abnormal when the number of the samples is less than a set value, classifying the samples again after removing the samples, and performing an effective index calculation formula according to an equation (1) when the number of the clusters is k:
wherein S is j And C j Respectively a jth classified curve and a corresponding clustering center; c k1 And C k2 Two different clustering centers when the classification number is k; n is a radical of s Is the number of data in the data set.
4. The transformer load prediction method according to claim 1, wherein: in the step f, from various types of typical samples obtained by cluster analysis, each type of sample is randomly selected and added into a neural network training set, so that all the typical types of samples are ensured to be included.
5. The transformer load prediction method according to claim 1, wherein: and g, in order to reduce the number of nodes of the ANN and improve the calculation speed, reducing the dimension of load sample data to obtain 24-point integral point load data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210718631.1A CN115049136A (en) | 2022-06-23 | 2022-06-23 | Transformer load prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210718631.1A CN115049136A (en) | 2022-06-23 | 2022-06-23 | Transformer load prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115049136A true CN115049136A (en) | 2022-09-13 |
Family
ID=83162626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210718631.1A Pending CN115049136A (en) | 2022-06-23 | 2022-06-23 | Transformer load prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115049136A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116029457A (en) * | 2023-02-21 | 2023-04-28 | 南方电网调峰调频发电有限公司鲁布革水力发电厂 | Electric power market electric energy coordination optimization method and system based on load data analysis |
-
2022
- 2022-06-23 CN CN202210718631.1A patent/CN115049136A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116029457A (en) * | 2023-02-21 | 2023-04-28 | 南方电网调峰调频发电有限公司鲁布革水力发电厂 | Electric power market electric energy coordination optimization method and system based on load data analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110991786B (en) | 10kV static load model parameter identification method based on similar daily load curve | |
| CN105117602B (en) | A kind of metering device running status method for early warning | |
| CN111105104A (en) | Short-term power load prediction method based on similar day and RBF neural network | |
| CN109492748B (en) | Method for establishing medium-and-long-term load prediction model of power system based on convolutional neural network | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN111008726B (en) | Class picture conversion method in power load prediction | |
| CN108805213B (en) | Power load curve double-layer spectral clustering method considering wavelet entropy dimensionality reduction | |
| CN112836604A (en) | Rolling bearing fault diagnosis and classification method, system and equipment based on VMD-SSAE and storage medium thereof | |
| CN111126868B (en) | Road traffic accident occurrence risk determination method and system | |
| CN106656357B (en) | Power frequency communication channel state evaluation system and method | |
| CN114676742A (en) | Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network | |
| CN112529638B (en) | Service demand dynamic prediction method and system based on user classification and deep learning | |
| CN112966871A (en) | Traffic jam prediction method and system based on convolution long-short term memory neural network | |
| CN112418476A (en) | Ultra-short-term power load prediction method | |
| CN105678406A (en) | Short-term load prediction method based on cloud model | |
| CN111882114B (en) | Short-time traffic flow prediction model construction method and prediction method | |
| CN115049136A (en) | Transformer load prediction method | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN114880538A (en) | Attribute graph community detection method based on self-supervision | |
| CN114500335A (en) | SDN network flow control method based on fuzzy C-means and mixed kernel least square support vector machine | |
| CN112149870B (en) | Pig house ammonia concentration combination prediction method based on ISODATA clustering and Elman neural network | |
| CN112950048A (en) | National higher education system health evaluation based on fuzzy comprehensive evaluation | |
| CN112836432A (en) | Indoor particle suspended matter concentration prediction method based on transfer learning | |
| CN112819087A (en) | Effluent BOD sensor abnormity detection method based on modular neural network | |
| CN110619422A (en) | Intelligent station passenger flow condition prediction method and system |
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 |