CN113762591B - Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning - Google Patents
Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning Download PDFInfo
- Publication number
- CN113762591B CN113762591B CN202110820888.3A CN202110820888A CN113762591B CN 113762591 B CN113762591 B CN 113762591B CN 202110820888 A CN202110820888 A CN 202110820888A CN 113762591 B CN113762591 B CN 113762591B
- Authority
- CN
- China
- Prior art keywords
- user
- gru
- data
- prediction
- short
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 230000005611 electricity Effects 0.000 claims abstract description 78
- 239000013598 vector Substances 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 26
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 25
- 230000007246 mechanism Effects 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 33
- 238000004590 computer program Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000010219 correlation analysis Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 230000015654 memory Effects 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 230000008485 antagonism Effects 0.000 claims description 4
- 230000006399 behavior Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 9
- 238000012545 processing Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 238000007781 pre-processing Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000000546 chi-square test Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
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
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses a method and a system for predicting short-term electric quantity of a user based on GRU and multi-core SVM countermeasure learning, wherein the method comprises the following steps: acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted; the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps: extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data; training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector; and taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user. According to the invention, the prediction performance is improved through the mutual game learning of the generator and the discriminator, and the accuracy of electric quantity prediction is improved.
Description
Technical Field
The invention relates to the technical field of intelligent electricity utilization, in particular to a short-term electricity quantity prediction method and system based on GRU and multi-core SVM countermeasure learning.
Background
The power consumption is accurately predicted, so that the accurate and reliable operation of the power grid system can be ensured, the resource waste in the power grid dispatching process is avoided, and meanwhile, the method is also beneficial to making a more economic power generation plan.
The traditional electric quantity prediction method mainly comprises a regression analysis method, a time sequence analysis method and the like. Although these methods are widely used in some applications in the power industry, they still have the disadvantage of being insufficiently considered for some uncertainty factors, and at the same time fail to make good use of the information of the sequence data. In order to improve the prediction performance, the intelligent prediction method (such as an artificial neural network, a support vector machine prediction model, a deep learning prediction method and the like) is gradually applied to electric quantity prediction research, and has the advantage of being capable of well mining the association relation between various influencing factors and electric quantity. Currently, most studies are predictive modeling based on electricity usage sequences, with recurrent neural networks (Recurrent Neural Network, RNN) being widely used. Long Short-Term Memory (LSTM) and gated loop units (Gated Recurrent Unit, GRU) are variants of RNN that can effectively address the deficiencies of Long-Term dependencies in RNN models, which are a relatively popular model for use in electrical quantity prediction studies. In addition, convolutional neural networks (Convolutional Neural Networks, CNN) have also been applied in load prediction studies. The inventors have found that although these methods have achieved good results, they ignore the potential correlations between the interior of the power sequences and do not fully exploit the existing information.
Disclosure of Invention
Based on the above problems, the first aspect of the present invention provides a method and a system for predicting short-term electric quantity of a user based on improved GRU and multi-core support vector machine (multi-core support vector machine) learning countermeasure. Based on main characteristic factors of electricity consumption, a bidirectional GRU and a multi-head attention mechanism are combined to serve as a generator and a multi-core SVM as a discriminator, short-term electricity prediction of a user is realized based on a generated countermeasure network model, prediction performance is improved through mutual game learning of the generator and the discriminator, and accuracy of electricity prediction is improved.
To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
the short-term electric quantity prediction method for the user based on GRU and multi-core SVM antagonism learning is characterized by comprising the following steps:
acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data;
training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector;
and taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user.
Further, after the user power consumption data and the power consumption related data are obtained, data cleaning is further performed, including deleting repeated data, complementing missing data and deleting error data.
Further, influence factor features and weights thereof are extracted based on a gray correlation analysis method.
Further, extracting influence factor features based on gray correlation analysis includes:
calculating feature correlation statistics by using a gray correlation analysis method;
and setting a correlation statistic threshold as a preset correlation threshold, and screening out characteristics related to the electricity consumption behavior of the user as influence factor characteristics.
Further, training the generator of the bi-directional GRU and multi-headed attentiveness mechanism-build includes:
based on the influence factor characteristics and the weights thereof, acquiring context information through bidirectional GRU learning;
performing multiple times of self-attention calculation, splicing calculation results of each time, and finally obtaining multiple-head attention scores through a linear mapping function to obtain a comprehensive vector of the power consumption of a user;
and calculating a loss function according to the softmax prediction function by adopting a comprehensive vector of the electricity consumption of the user, training learning parameters of the bidirectional GRU by adopting a back propagation algorithm, and completing training of the generator.
Further, obtaining context information through bi-directional GRU learning includes:
mapping the influence factor characteristic sequence vector into a low-dimensional vector set;
respectively learning the low-dimensional vector set through the forward GRU and the backward GRU;
and splicing the features obtained by the bidirectional GRU learning through the splicing function to obtain the context information.
Further, a root mean square error is calculated according to the actual power consumption and the generated power consumption predicted value, and the root mean square error is used as an objective function to update parameters of the generator.
One or more embodiments provide a user short-term power prediction system based on GRU and multi-core SVM challenge learning, comprising:
a data acquisition module configured to: acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
a power consumption prediction module configured to: the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data;
training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector;
and taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user.
One or more embodiments provide an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, which when executed implements the GRU and multi-core SVM-based user short-term power prediction method.
One or more embodiments provide a computer readable storage medium having a computer program stored thereon, wherein the program when executed by a processor implements the GRU and multi-core SVM challenge learning based method for short-term power prediction of a user.
The technical scheme has the following beneficial effects:
the invention is based on historical user electricity consumption data and electricity consumption related data, combines a bidirectional GRU and a multi-head attention mechanism as a generator, uses a multi-core SVM as a discriminator, realizes short-term electricity prediction of a user based on a generated countermeasure network model, and improves prediction performance through mutual countermeasure learning of the generator and the discriminator.
Aiming at the characteristics of time sequence and long-term dependence of the historical electricity consumption data of the user, the invention adopts the bidirectional GRU to predict the future electricity consumption condition; in addition, in order to capture the internal structure of the electric quantity sequence, the dependency relationship between the internal data information of the sequence is learned, the thought of a multi-head attention mechanism is introduced, the characteristic information is represented from different dimensions and different subspaces, and the accuracy of model prediction is improved.
The invention provides the method for adjusting the generator by adopting root mean square error of real data and generated data as a part of an objective function, thereby reducing the error probability of the model and improving the stability of the model.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flowchart illustrating a method for predicting short-term power of a user based on GRU and multi-core SVM challenge learning according to one or more embodiments of the present invention;
FIG. 2 is a flow chart of a method for preprocessing data according to one or more embodiments of the present invention;
FIG. 3 is an overall schematic diagram of a method for user short-term power prediction based on improved GRU and multi-core SVM countermeasure learning, provided by one or more embodiments of the invention;
FIG. 4 is a schematic diagram of a multi-headed attention mechanism provided by one or more embodiments of the present invention;
fig. 5 is an effect diagram of a method for predicting short-term electric quantity of a user based on GRU and multi-core SVM challenge learning according to one or more embodiments of the present invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
As shown in fig. 1, the method for predicting the short-term electric quantity of the user based on the improved GRU and the multi-core SVM countermeasure learning according to the embodiment includes the following steps:
step 1: acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
step 2: the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
s1: acquiring historical electricity utilization data and electricity utilization related data, and extracting influence factor characteristics and weights thereof; the method specifically comprises the following steps:
A. the collection of related electricity consumption data is performed on mass electric power information, the related electricity consumption data comprises historical electricity consumption data and electricity consumption related data, the electricity consumption related data comprises weather data, holiday data and the like, and in the embodiment, the weather data is taken as an example for illustration. And carrying out data preprocessing on the acquired electricity consumption data and weather data, wherein the data preprocessing comprises data cleaning, missing data completion, data definition and storage.
Specifically, based on electricity consumption data acquired in an electricity consumption data acquisition system of a provincial power grid company in China, the data of 24 time points are acquired every hour, wherein the data comprise 334656 pieces of electricity consumption data of 996 users from 1 day in 2019 to 14 days in 2019. It is to be noted here that the present embodiment uses the data collected at each time point as one piece of electricity consumption data. In addition, weather data of corresponding cities at corresponding times are obtained from the national weather data service center website.
B. And (3) carrying out standardization processing on the massive electric quantity data, carrying out characteristic selection affecting the electric quantity on the weather condition based on a gray correlation analysis method (Grey Relation Analysis, GRA), eliminating redundancy of the characteristics and extracting the characteristics with high correlation degree with the affecting electric quantity.
The method comprises the steps of preprocessing acquired data, including missing value processing and data normalization, and performing simple correlation analysis by adopting chi-square test, so that the arbitrary selection of repeated influence factors and factors is eliminated, the complexity of the problem is reduced, the redundancy of the characteristics is eliminated, and more information variables are selected to improve the accuracy and efficiency of a prediction model.
In the process of predicting the electricity consumption of the user, weather conditions have important influence on the prediction of the electricity consumption. And aiming at the acquired electric quantity data sample, according to a gray correlation analysis method, the result of distinguishing feature selection is shown as follows. Table 1 shows the calculation of the relevant statistics between weather factors and user behavior by the GRA algorithm and ordering the statistics from high to low, a threshold (=0.5) is set to exclude some non-informative features. In the experiment, it can be concluded that two weather factors, namely temperature and humidity, have a larger influence on the electricity consumption prediction.
TABLE 1 weather factor correlation analysis
| Factors of | Correlation degree |
| Maximum air temperature | 0.9819 |
| Minimum air temperature | 0.9707 |
| Humidity of the water | 0.9674 |
| Average temperature | 0.9199 |
| Total precipitation | 0.8651 |
| Wind speed | 0.8470 |
| Air pressure | 0.6895 |
B. Specifically, as shown in fig. 2, the general generation process of weather factor analysis based on the gray correlation analysis method (Grey Relation Analysis, GRA) in the step B is as follows:
B1. normalizing the historical electric quantity data and the corresponding weather data, and normalizing the original data X by adopting a min-max normalization method, wherein the value of the characteristic data is [0,1];
B2. assume that the power consumption reference sequence after each characteristic data in the original sample is subjected to mean change is X 0 ＝(x 0 (1)，x 0 (2)，......，x 0 (n)) and setting the historical power data including various factor characteristics as a comparison sequence X k ＝(x k (1)，x k (2)，......，x k (n)), i=1, 2,..m. Then X is 0 And X is k The calculation formula of the correlation coefficient of (2) is as follows:
in delta i (k) Represents x i For x 0 Correlation coefficients over K data. ρ represents a resolution coefficient (ρ=0.5);
B3. calculating the association degree theta of each factor and the electricity consumption of the user
Finally, the main feature set influencing the medical migration can be selected according to the sorting degree by sorting the relevance theta, and the screened feature set is output to provide basic support for the construction of a later electric quantity prediction model.
S2: training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector; specifically, a bidirectional GRU is constructed based on the acquired factor characteristic set and the corresponding weight thereof, so that the electricity consumption information of the user is learned from the front direction and the back direction; then, a generator is constructed by combining a multi-head attention mechanism; updating weight values of the power consumption sequence information state prediction data of each user in the bidirectional GRU, and outputting a power consumption state prediction vector of the user; constructing a softmax prediction function by using the user power consumption state prediction vector; and calculating a loss function of the output value of the softmax prediction function, training learning parameters of the bidirectional GRU by adopting a back propagation algorithm, and completing training of the generator.
The step S2 specifically comprises the following steps:
C. and C, based on the data preprocessed in the step B and the acquired characteristic factor set, adopting a bidirectional GRU and multi-head attention combination to construct a generator, and realizing electric quantity prediction, as shown in figure 3.
C1. Base groupConstructing an input sequence { x } of a model from the obtained set of factors affecting the power consumption of the user 1 ，x 2 ，...，x t Through the operation of an Embedding layer, mapping the sequence vector with high-dimensional sparseness into a vector set { e with low-dimensional density } 1 ，e 2 ，...e t }。
e i ＝W T x i (i＝1，2，...，t) (3)
Wherein W is E R |t|*d The feature matrix is represented, t represents the length of the sequence, and K represents the Embedding dimension of the Embedding layer.
C2. Based on the acquired low-dimensional dense vector set, for each time t, the GRU uses the input e t And previous state h t-1 Calculate h t The following are provided:
r t ＝σ(W r e t +U r h t-1 ) (4)
n t ＝σ(W π e t +U π h t-1 ) (5)
wherein h is t , r t and pi t The hidden state of the d-dimension, reset gate and update gate, respectively. W (W) r ,W π ,W c And U r ,U π U is a parameter of the GRU. Sigma is a sigmoid function.
The Bi-GRU consists of a forward GRU and a backward GRU, and the context information of the sequence data is better acquired through learning in the front direction and the back direction. Finally, the final hidden state H is obtained through Concat function splicing t The following is shown:
wherein,representing hidden state learned by forward GRU +.> Representing the hidden state learned from the opposite direction +.>
C3. In order to capture the internal structure of the sequence, the dependency relationship between the internal data information of the sequence is learned, a multi-head attention mechanism is introduced, multiple self-attention calculations are carried out, each calculation result is spliced, and finally the multi-head attention score is obtained through a linear mapping function, as shown in fig. 4.
First, self-attention score is performedThe calculation is as follows:
wherein H= { H 1 ，H 2 ，....，H t The matrix is composed of output vectors at all times of the bidirectional GRU layer. W (W) l ，γ，Representing the parameter vector.
The characteristic value of single attention output can be obtained through self-attention calculation as follows:
then, K times of calculation are performed using formulas (9) to (11). Will result H * Splicing and linear mapping are carried out, and a final result is obtained:
C4. finally, the attention characteristic sequence obtained through the multi-head attention mechanism is input into a Softmax layer to conduct electricity consumption prediction, and the prediction result is as follows:
Y＝Softmax(W r H final +b r ) (13)
s3: and taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user. Specifically, based on the obtained prediction vector of the power consumption state of the user and the real data, constructing a discriminator by utilizing a multi-core SVM, and discriminating whether the result predicted by the generator belongs to the real data or not; based on the discrimination result obtained by the discriminator, the data information in the generator is fed back and updated, and the prediction of the power consumption of the user is continuously perfected by mutually resisting the data weight value in the continuous optimization model.
The step S3 specifically comprises the following steps:
D. c, based on the predicted vector of the user power consumption state obtained in the step C, combining with the real data, constructing a discriminator by utilizing a multi-core SVM, and discriminating whether the predicted result of the generator belongs to the real data.
The multi-core SVM is an expansion of a single-core SVM, and the aim is to determine the optimal combination of M kernel functions so as to maximize the distance, and the method can be represented by the following optimization problem:
wherein Δ= { θ∈r+|θ T e M =1 } represents coefficients of convex combinations of M kernel functions; e, e M Representing vectors where M elements are all 1;represents the final kernel function, where k j (. Cndot. ) is the j-th kernel function.
The Lagrangian multiplier method is used for the conversion of equation (14) to the following optimized form:
wherein K is j ∈R N×N ,Ω＝{α|α∈[0，C] N - α is the lagrangian multiplier; "×" is defined as the dot product of the vectors.
E. Based on the discrimination result obtained in the step D, the data information in the generator is fed back and updated, and the data weight value in the model is continuously optimized through mutual antagonism, so that the power consumption prediction of the user is continuously perfected. Adopting cross entropy as a loss function, if y is a real class distribution, defining the loss function as follows:
wherein S is positive And S is negative Positive sample data and negative sample data, respectively. P (P) discriminator (Y t X represents the sample (Y) t X) probability of belonging to real data.
In the model training process, the objective function of the generator is not related to the real data, and the parameters of the generator are adjusted only by the output of the discriminator. When the judging device judges that the error exists, the data generated by the generator is deviated, and the model is not stable enough. In this implementation, therefore, the mean square error of the real samples and the generated samples is taken as part of the generator objective function. When the judgement device is in error, the generator can be regulated by means of root mean square error, so that the error probability of the model can be reduced, and the stability of the model can be raised.
And carrying out electricity consumption prediction on a sample to be predicted, pushing a prediction result, comparing the result with actual data, and carrying out method analysis by adopting two evaluation indexes of root mean square error (Root Mean Squared Error, RMSE) and average absolute percentage error (MeanAbsolute Percentage Error, MAPE). The initial settings of the model parameters in this embodiment are shown in table 2.
Wherein y is t Representing the true value at time t; y' t Representing a predicted value; n represents the number of data.
TABLE 2 model parameter settings
| Parameter index | Quantity of |
| Embedding dimensions | 100 |
| Two-way GRU hidden layer node number | 64 |
| Multi-head attention number | 8 |
| Learning rate | 0.001 |
| Dropout | 0.5 |
| Number of cores of multi-core SVM | 5 |
| Iteration number (Epoch) | 100 |
Table 3 describes a comparison of method performance in electricity consumption prediction: based on the results in table 3, the prediction performance of the electricity consumption prediction model proposed in this embodiment is superior to other prediction methods.
TABLE 3 comparison of Performance of different prediction methods
Example two
The embodiment aims to provide a short-term electric quantity prediction system based on GRU and multi-core SVM antagonism learning.
Based on the above object, the present embodiment provides a short-term electricity prediction system based on GRU and multi-core SVM challenge learning, including:
a data acquisition module configured to: acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
a power consumption prediction module configured to: the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data;
training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector;
and taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user.
Example III
An object of the present embodiment is to provide an electronic apparatus.
In view of the above, the present embodiment provides an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method according to the first embodiment when executing the program.
Example IV
An object of the present embodiment is to provide a computer-readable storage medium.
In view of the above, the present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method as described in embodiment one.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (8)
1. The short-term electric quantity prediction method for the user based on GRU and multi-core SVM antagonism learning is characterized by comprising the following steps:
acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data;
training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector;
training the generator of the bidirectional GRU and multi-headed attention mechanism construct includes:
based on the influence factor characteristics and the weights thereof, acquiring context information through bidirectional GRU learning;
performing multiple times of self-attention calculation, splicing calculation results of each time, and finally obtaining multiple-head attention scores through a linear mapping function to obtain a comprehensive vector of the power consumption of a user;
calculating a loss function according to a softmax prediction function by adopting a comprehensive vector of the electricity consumption of a user, training learning parameters of a bidirectional GRU by adopting a back propagation algorithm, and finishing training of a generator;
taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user;
and calculating the root mean square error according to the actual power consumption and the generated power consumption predicted value, and updating parameters of the generator by taking the root mean square error as an objective function.
2. The method for predicting short-term electric quantity of a user based on GRU and multi-core SVM countermeasure learning according to claim 1, wherein after acquiring the user power consumption data and the power consumption related data, data cleaning is further performed, including deleting duplicate data, complementing missing data and deleting error data.
3. The method for predicting short-term electric quantity of a user based on GRU and multi-core SVM countermeasure learning according to claim 1, wherein the influence factor features and weights thereof are extracted based on a gray correlation analysis method.
4. The method for predicting short-term power of a user based on GRU and multi-core SVM challenge learning of claim 3, wherein extracting the influence factor features based on gray correlation analysis comprises:
calculating feature correlation statistics by using a gray correlation analysis method;
and setting a correlation statistic threshold as a preset correlation threshold, and screening out characteristics related to the electricity consumption behavior of the user as influence factor characteristics.
5. The method of short-term power prediction for a user based on GRU and multi-core SVM challenge learning of claim 1, wherein obtaining context information through bi-directional GRU learning comprises:
mapping the influence factor characteristic sequence vector into a low-dimensional vector set;
respectively learning the low-dimensional vector set through the forward GRU and the backward GRU;
and splicing the features obtained by the bidirectional GRU learning through the splicing function to obtain the context information.
6. A user short-term power prediction system based on GRU and multi-core SVM challenge learning, comprising:
a data acquisition module configured to: acquiring historical electricity utilization data and current electricity utilization related data of a user to be predicted;
a power consumption prediction module configured to: the power consumption prediction is carried out by adopting a pre-trained user short-term power prediction model, wherein the training method of the user short-term power prediction model comprises the following steps:
extracting influence factor characteristics and weights thereof based on historical electricity consumption data and electricity consumption related data;
training a generator constructed by the bidirectional GRU and the multi-head attention mechanism based on the influence factor characteristics and the weights thereof, and outputting a user power consumption state prediction vector;
training the generator of the bidirectional GRU and multi-headed attention mechanism construct includes:
based on the influence factor characteristics and the weights thereof, acquiring context information through bidirectional GRU learning;
performing multiple times of self-attention calculation, splicing calculation results of each time, and finally obtaining multiple-head attention scores through a linear mapping function to obtain a comprehensive vector of the power consumption of a user;
calculating a loss function according to a softmax prediction function by adopting a comprehensive vector of the electricity consumption of a user, training learning parameters of a bidirectional GRU by adopting a back propagation algorithm, and finishing training of a generator;
taking the comprehensive vector of the user electricity consumption and the real data as input, and updating parameters of the generator based on the judging result of the multi-core SVM discriminator to obtain a short-term electricity prediction model of the user;
and calculating the root mean square error according to the actual power consumption and the generated power consumption predicted value, and updating parameters of the generator by taking the root mean square error as an objective function.
7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the GRU and multi-core SVM challenge learning based user short term power prediction method of any of claims 1-5 when the program is executed by the processor.
8. A computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a method for short-term power prediction of a user based on GRU and multi-core SVM challenge learning as claimed in any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110820888.3A CN113762591B (en) | 2021-07-20 | 2021-07-20 | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110820888.3A CN113762591B (en) | 2021-07-20 | 2021-07-20 | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113762591A CN113762591A (en) | 2021-12-07 |
| CN113762591B true CN113762591B (en) | 2024-04-05 |
Family
ID=78787793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110820888.3A Active CN113762591B (en) | 2021-07-20 | 2021-07-20 | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113762591B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115808944B (en) * | 2023-02-09 | 2023-06-02 | 国能大渡河枕头坝发电有限公司 | Stator temperature rise test control method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508642A (en) * | 2018-10-17 | 2019-03-22 | 杭州电子科技大学 | Ship monitor video key frame extracting method based on two-way GRU and attention mechanism |
| CN109711620A (en) * | 2018-12-26 | 2019-05-03 | 浙江大学 | A kind of Short-Term Load Forecasting Method based on GRU neural network and transfer learning |
| CN110516831A (en) * | 2019-06-18 | 2019-11-29 | 国网（北京）节能设计研究院有限公司 | A kind of short-term load forecasting method based on MWOA algorithm optimization SVM |
| CN110942205A (en) * | 2019-12-05 | 2020-03-31 | 国网安徽省电力有限公司 | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM |
| CN112270454A (en) * | 2020-11-19 | 2021-01-26 | 国网北京市电力公司 | Method and device for predicting short-term load of power system under influence of extreme factors |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11449744B2 (en) * | 2016-06-23 | 2022-09-20 | Microsoft Technology Licensing, Llc | End-to-end memory networks for contextual language understanding |
| US11665576B2 (en) * | 2018-08-10 | 2023-05-30 | Verizon Patent And Licensing Inc. | Systems and methods for wireless low latency traffic scheduler |
-
2021
- 2021-07-20 CN CN202110820888.3A patent/CN113762591B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508642A (en) * | 2018-10-17 | 2019-03-22 | 杭州电子科技大学 | Ship monitor video key frame extracting method based on two-way GRU and attention mechanism |
| CN109711620A (en) * | 2018-12-26 | 2019-05-03 | 浙江大学 | A kind of Short-Term Load Forecasting Method based on GRU neural network and transfer learning |
| CN110516831A (en) * | 2019-06-18 | 2019-11-29 | 国网（北京）节能设计研究院有限公司 | A kind of short-term load forecasting method based on MWOA algorithm optimization SVM |
| CN110942205A (en) * | 2019-12-05 | 2020-03-31 | 国网安徽省电力有限公司 | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM |
| CN112270454A (en) * | 2020-11-19 | 2021-01-26 | 国网北京市电力公司 | Method and device for predicting short-term load of power system under influence of extreme factors |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113762591A (en) | 2021-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109754113B (en) | Load prediction method based on dynamic time warping and long-and-short time memory | |
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN111738512A (en) | Short-term power load prediction method based on CNN-IPSO-GRU hybrid model | |
| CN110751318B (en) | Ultra-short-term power load prediction method based on IPSO-LSTM | |
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN108921339B (en) | Quantile regression-based photovoltaic power interval prediction method for genetic support vector machine | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN112884008B (en) | Prediction evaluation method and device for running state of power information acquisition system | |
| CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
| CN113762387B (en) | Multi-element load prediction method for data center station based on hybrid model prediction | |
| CN113344288B (en) | Cascade hydropower station group water level prediction method and device and computer readable storage medium | |
| CN109492748A (en) | A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks | |
| CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
| CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
| CN111428766A (en) | Power consumption mode classification method for high-dimensional mass measurement data | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN114119273A (en) | Park comprehensive energy system non-invasive load decomposition method and system | |
| CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network | |
| CN117313795A (en) | Intelligent building energy consumption prediction method based on improved DBO-LSTM | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN117113086A (en) | Energy storage unit load prediction method, system, electronic equipment and medium | |
| Wang et al. | Cloud computing and extreme learning machine for a distributed energy consumption forecasting in equipment-manufacturing enterprises | |
| Gao et al. | Establishment of economic forecasting model of high-tech industry based on genetic optimization neural network | |
| CN112836876B (en) | Power distribution network line load prediction method based on deep learning | |
| Lv et al. | Research on short-term load forecasting approach for smart grid |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |