CN108921358B - Prediction method, prediction system and related device of power load characteristics - Google Patents
Prediction method, prediction system and related device of power load characteristics Download PDFInfo
- Publication number
- CN108921358B CN108921358B CN201810777121.5A CN201810777121A CN108921358B CN 108921358 B CN108921358 B CN 108921358B CN 201810777121 A CN201810777121 A CN 201810777121A CN 108921358 B CN108921358 B CN 108921358B
- Authority
- CN
- China
- Prior art keywords
- load
- prediction
- establishing
- model
- probability interval
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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 application provides a prediction method of power load characteristics, which comprises the following steps: performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set; establishing a training sample according to the load optimal feature set; establishing a Gaussian process regression model by using the training samples; constructing a load probability interval prediction model under a specific confidence level; and outputting a prediction result by using a load probability interval prediction model. The load probability interval prediction method based on the K-means feature extraction method and the DSOGPR are combined to construct the load probability interval prediction result under the specific confidence level, the dual-strategy online Gaussian process regression algorithm is achieved, the accurate prediction result can be obtained, potential safety hazards and economic losses caused by unreasonable power decision making are avoided, and the method has important significance. The application also provides a power load characteristic prediction system, a computer readable storage medium and a power load characteristic prediction terminal, which have the beneficial effects.
Description
Technical Field
The present disclosure relates to the field of power and power grid systems, and in particular, to a method and a system for predicting power load characteristics, a computer-readable storage medium, and a power load characteristic prediction terminal.
Background
The electric power industry is concerned with the national people and is closely related to the production and living activities of people, and the stable operation of an electric power system is a necessary condition for social development and economic development. The stability of a power system is maintained, and the balance of supply and demand of power is maintained as a basis, namely, the power generation amount of the system and the power load of a power consumer must be balanced in real time, and load prediction is a very important ring in the balance of supply and demand. If the predicted value of the load is lower than the actual value, the power consumption requirement of a user cannot be met, and the problem of power shortage is caused; if the predicted value of the load is higher than the actual value, the power is not stored in a large amount, which results in waste of power resources. Only accurate prediction of power can help the power industry to develop continuously and healthily.
The load prediction is to obtain the load estimation value at the future time according to the historical load data, the operation characteristics of the system, social conditions, natural conditions (temperature, humidity, and the like), economic indexes, and the like. How to effectively utilize the data and adopt a proper method for prediction to ensure the accuracy of prediction is a main problem to be solved by a load prediction technology.
With the power users of various random terminals such as distributed new energy power generation, photovoltaic power generation and grid connection of electric vehicles, uncertain factors and unstable factors of a power system gradually increase, greater risks are brought to power decision work, and new challenges are brought to power load prediction work. With the grid connection of renewable energy sources such as large-scale wind power and the like, the uncertainty of the power demand needs to be accurately reflected in the economic dispatching of the power system, however, the predictive value of the deterministic model is single, and more comprehensive load information cannot be provided for a power decision maker to refer to.
The short-term load prediction method considering the uncertain requirement of the power demand is an effective tool for helping the power grid to guarantee stable operation and reasonable planning, and can provide necessary information for the operation and scheduling of the power system.
Content of application
The purpose of the present application is to provide a power load characteristic prediction method, a power load characteristic prediction system, a computer-readable storage medium, and a power load characteristic prediction terminal, which solve the problem of low accuracy of the existing power load characteristic prediction.
In order to solve the above technical problem, the present application provides a method for predicting a power load characteristic, which has the following specific technical scheme:
performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
establishing a training sample according to the load optimal feature set; wherein, the samples in the training samples are input variable matrixes and corresponding target values;
establishing a Gaussian process regression model by using the training samples;
according to the Gaussian process regression model, a load probability interval prediction model under a specific confidence level is built;
and inputting the data in the load optimal characteristic set, and outputting a prediction result by using the load probability interval prediction model.
The method for selecting the characteristics of the historical load data and the new load data and outputting the load optimal characteristic set by using the K-Means characteristic extraction method comprises the following steps:
constructing a candidate characteristic set for the historical load data by using a K-Means characteristic extraction method;
performing feature clustering on the candidate feature set to form a first candidate feature set;
adding new load data into the first candidate feature set to form a second candidate feature set;
and performing online feature classification and feature selection on the second candidate feature set, and outputting a load optimal feature set.
Wherein, still include:
updating the load-optimized feature set each time the new load data is added to the first candidate feature set.
After the training sample is established according to the load optimal feature set, the method further comprises the following steps:
and updating the training samples by using a sliding time window technology, and ensuring that the number of the training samples is kept unchanged during updating.
According to the Gaussian process regression model, a load probability interval prediction model under a specific confidence level is constructed, and the method comprises the following steps:
establishing a relational expression between an input matrix and an output matrix according to the Gaussian process regression model;
establishing a finite set conforming to joint Gaussian distribution according to the relational expression;
determining a kernel matrix in the finite set, determining a distribution function of the target values;
determining the edge distribution of the distribution function, determining a hyper-parameter set according to the edge distribution, and obtaining a log-likelihood function of the Gaussian process regression model;
calculating the joint distribution of the input samples and the input variable matrix;
establishing a prediction model of the input sample according to the joint distribution;
and constructing a load probability interval prediction model under a specific confidence level according to a preset confidence level and the prediction model.
After inputting the data in the load optimal feature set and outputting a prediction result by using the load probability interval prediction model, the method further comprises the following steps:
and updating the model parameters of the load probability interval prediction model by using a recursive least square method based on forgetting factors.
Wherein, still include:
and selecting one or more of the predicted clearance coverage rate, the normalized average width of the predicted interval, the comprehensive index and the average absolute percentage error to carry out performance evaluation on the load probability interval prediction model.
The present application also provides a system for predicting a characteristic of an electrical load, the system comprising:
the data selection module is used for performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
the sample establishing module is used for establishing a training sample according to the load optimal feature set; wherein, the samples in the training samples are input variable matrixes and corresponding target values;
the Gaussian model establishing module is used for establishing a Gaussian process regression model by using the training sample;
the prediction model establishing module is used for establishing a load probability interval prediction model under a specific confidence level according to the Gaussian process regression model;
and the prediction module is used for inputting the data in the load optimal characteristic set and outputting a prediction result by using the load probability interval prediction model.
The present application also provides a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the prediction method as described above.
The application also provides a power load characteristic prediction terminal which comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor realizes the steps of the prediction method when calling the computer program in the memory.
The application provides a prediction method of power load characteristics, which comprises the following steps: performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set; establishing a training sample according to the load optimal feature set; establishing a Gaussian process regression model by using the training samples; wherein, the samples in the training samples are input variable matrixes and corresponding target values; according to the Gaussian process regression model, a load probability interval prediction model under a specific confidence level is built; and inputting the data in the load optimal characteristic set, and outputting a prediction result by using the load probability interval prediction model.
The application provides a dual-strategy online Gaussian process regression algorithm, which is combined with a K-means-based feature extraction method to select model input variables. Firstly, screening an optimal input variable set from a candidate feature set by using a KFS method, and then constructing a load probability interval prediction result under a specific confidence level by using Double Strategy Online Gaussian Process Regression (DSOGPR). With the power users of various random terminals such as distributed new energy power generation, photovoltaic power generation and grid connection of electric vehicles, uncertain factors and unstable factors of a power system gradually increase, greater risks are brought to power decision work, and new challenges are brought to power load prediction work. The method and the device have important significance in avoiding potential safety hazards and economic losses caused by unreasonable power decision-making, and the accurate load prediction result can play a role in stable operation, reasonable planning and the like of the power grid. The application also provides a power load characteristic prediction system, a computer readable storage medium and a power load characteristic prediction terminal, which have the beneficial effects and are not repeated herein.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a method for predicting a power load characteristic according to an embodiment of the present disclosure;
FIG. 2 is a flow chart of a method for feature extraction using K-Means according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a system for predicting a power load characteristic according to an embodiment of the present disclosure.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. 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 application.
In order to make the description of the steps of the present application clearer, the following description will use mathematical expressions corresponding to the steps, but the mathematical expressions corresponding to the steps should not be understood as a unique expression corresponding to the steps, and only serve as one possible embodiment among the embodiments of the present application.
Referring to fig. 1, fig. 1 is a flowchart of a method for predicting a power load characteristic according to an embodiment of the present disclosure, where the method includes:
s101: performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
this step is intended to make the subsequent predictive model more accurate on the basis of both types of data, by using the historical load data and the new load data for processing and selection. Because the power load prediction process is dynamically changed, under the dynamic characteristic space, the characteristic selection of the historical load data and the new load data stream in the whole scene is considered, and the accuracy of the prediction result is better ensured. In order to solve the problem, the method provides a feature extraction method based on K-Means (also called KFS). Meanwhile, the historical load data and the new load data can be subjected to dimensionality reduction by using a K-Means feature extraction method. The load data generally comprises multiple factors such as time, operation characteristics, natural conditions and the like, and after the K-Means characteristics are extracted, the factors which have no influence or low influence in the power load prediction process can be eliminated, one or more factors which have the largest influence on the power load prediction result are obtained, and the operation difficulty is reduced.
The specific steps of the K-Means feature extraction method are not limited in the application, and the method can be based on a KFS method, and can also be improved on the basis of the KFS method. A specific feature extraction process is proposed herein, and referring to fig. 2, fig. 2 is a flowchart of a method for extracting features by using K-Means provided in an embodiment of the present application, and the operation process may be as follows:
s1011: constructing a candidate characteristic set for the historical load data by using a K-Means characteristic extraction method;
s1012: performing feature clustering on the candidate feature set to form a first candidate feature set;
s1013: adding new load data into the first candidate feature set to form a second candidate feature set;
s1014: and performing online feature classification and feature selection on the second candidate feature set, and outputting a load optimal feature set.
It should be noted that, the processes of feature clustering, online feature classification, feature selection, and the like all require a person skilled in the art to perform corresponding setting according to the actual requirements in the power load prediction process, and are not specifically limited herein.
In addition, when new load data is added into the first candidate feature set, the load optimal feature set can be updated, and the real-time performance and the accuracy of a prediction result can be guaranteed.
Of course, other feature extraction processes based on the K-Means feature extraction method may exist, and are not described herein by way of example, and all of them should be considered within the scope of the present application.
S102: establishing a training sample according to the load optimal feature set;
wherein, samples in the training samples are input variable matrixes and corresponding target values;
the samples in the training samples are established according to the load optimal feature set obtained in S101. For example, for a given set of N sample dataWherein: x is the number ofi∈RdAs a matrix of input variables, ynIs its corresponding target value. Training sample given a time windowAnd online learning sampleAccording to the definition of the Gaussian process, the relation between the input matrix and the output matrix can be established;
s103: establishing a Gaussian process regression model by using the training samples;
establishing a gaussian process regression model is a mature prior art and is not described herein.
S104: according to a Gaussian process regression model, constructing a load probability interval prediction model under a specific confidence level;
the step aims to establish a final load probability interval prediction model according to a Gaussian process regression model. The present application provides a specific load probability interval prediction model building process, and certainly, other building processes may be provided, which are not limited herein. The specific establishment process may be as follows:
s1041: establishing a relational expression between an input matrix and an output matrix according to the Gaussian process regression model;
yn=f(xn)+εn (1)
where f is a function defined under the data set D, εnTo obey distributionIndependent white gaussian noise.
S1042: establishing a finite set conforming to joint Gaussian distribution according to the relational expression;
a finite set f ═ f (x)1),f(x2),…,f(xN) Constitute a set of random process variables and have a joint gaussian distribution, i.e.:
p(f|x1,x2,…,xN)~(0,K) (2)
s1043: determining a kernel matrix in the finite set, determining a distribution function of the target values;
(2) wherein K is a nuclear matrix whose elements Kij=k(xi,xj) And k (·) is a kernel function. In this embodiment, a squared exponential covariance function (SE) is selected as a kernel function, which is defined as:
in the formula (I), the compound is shown in the specification,is the signal equation of the kernel function, and is a hyperparametric symmetric matrix,is the variance of the noise.
According to Bayesian theory, GP (Gaussian regression method, GPR hereafter) at given data D0A prior function is established in the set, and under the condition of given f, the distribution obeyed by y, namely the distribution function of the target value is
S1044: determining the edge distribution of the distribution function, determining a hyper-parameter set according to the edge distribution, and obtaining a log-likelihood function of the Gaussian process regression model;
since the distribution (4) is an isotropic gaussian distribution, the edge distribution of y can be calculated by equation (5):
GP determines hyper-parameter set by finding edge distribution (6) during trainingThe log-likelihood function of the GP model can be derived:
then only the maximum value of the solution formula (6) is needed, the hyperparameter is initialized, the optimal hyperparameter is determined under the Bayes framework of the maximum likelihood, and the optimal hyperparameter can be obtained by solving the partial derivative of theta through the formula:
where Tr () represents a trace of the matrix.
S1045: calculating the joint distribution of the input samples and the input variable matrix;
to this end, a new input sample X is predicted using the previously established GP model, and a joint distribution p (X | X, X) is calculated, i.e.
Wherein y ═ y1,y2,…,yN,y*]TAnd K is a core matrix of (N +1) × (N + 1). Since the joint distribution also follows a gaussian distribution, equation (8) can be expressed as
Wherein k ═ k (x)1,x*),…,k(xN,x*)]T,k**=k(x*,x*)。
S1046: establishing a prediction model of the input sample according to the joint distribution;
by using gaussian condition distribution, the prediction distribution of the new sample x can be obtained, namely the prediction model:
s1047: and constructing a load probability interval prediction model under a specific confidence level according to a preset confidence level and the prediction model.
From equation (10), the prediction interval at a given confidence level α, i.e.
In the formula (10) is defined,
it should be noted that the above process is only a specific load probability interval prediction model establishment process, and those skilled in the art can also obtain other prediction model establishment processes similar to or based on the content provided by the present application, and all of them are within the scope of the present application.
S105: and inputting the data in the load optimal characteristic set, and outputting a prediction result by using the load probability interval prediction model.
After the load probability interval prediction model is obtained in S1046, the power load characteristic may be predicted according to the load probability interval prediction model.
In the prediction process, data subjected to K-Means feature extraction and dimensionality reduction in S101, namely data in the load optimal feature set, are used. The predicted validity period of the result is relatively short, typically one to two weeks.
According to the method, the optimal input variable set is screened by using an online feature extraction method based on kmeans feature clustering, an online learning technology is combined with a Gaussian process regression method, uncertainty hidden in load data is responded by updating model parameters, the influence caused by noise can be reduced by a new algorithm, and the method has good tracking capability and adaptability.
Meanwhile, on the background that the power data are gradually complete and rich, the method and the device make full use of the historical load data of the power users, analyze the power utilization regularity and the load form of each specific user, and combine a k-means feature extraction algorithm, so that the calculation time and the information storage amount are saved, and the power selling company can provide differentiated value-added services for the users with different power utilization characteristics conveniently. The method is convenient for the power selling company to design time-of-use electricity price packages, smooth regional load curves, sign interruptible load protocols and reduce deviation assessment risks, and provides a solid foundation.
Based on the foregoing embodiment, as a preferred embodiment, after S102, the method may further include:
and updating the training samples by using a sliding time window technology, and ensuring that the number of the training samples is kept unchanged during updating.
By using D0And establishing a Gaussian Process Regression (GPR) model. Updating training samples by using sliding time window technique, and changing M old data samples from D at intervals0Removing, adding M new samples to ensure D0Number of samples T0And is not changed.
Based on the foregoing embodiment, as a preferred embodiment, after S105, the method may further include:
and updating the model parameters of the load probability interval prediction model by using a recursive least square method based on forgetting factors.
DonlineIn the online learning stage, a is updated by using a Recursive least square method (RLS-FF) based on forgetting factorsnAnd CnI.e. the mean function and covariance function of the updated samples. Updating a with RLS-FFnAnd CnThe implementation process is as follows:
k*(n+1)=[k(x1,x*),…,k(xN,x*)]T (17)
wherein, PnTo transfer the matrix, en+1To prediction error, xin+1Are auxiliary parameters. Matrix PnThe updates of (2) are as follows:
wherein epsilonn+1As an auxiliary parameter, λnThe calculation formula of the forgetting factor is as follows:
νn+1=λn(νn+1) (22)
the embodiment aims to improve the prediction accuracy of the prediction method, and combines an online learning strategy and a Gaussian process regression method according to the historical load data of the power user, so that better prediction interval quality, narrower interval width and higher prediction accuracy are realized.
Based on any one of the above embodiments, as a preferred embodiment, the method may further include, after predicting:
and selecting one or more of the predicted clearance coverage rate, the normalized average width of the predicted interval, the comprehensive index and the average absolute percentage error to carry out performance evaluation on the load probability interval prediction model.
In order to qualitatively evaluate the performance of the prediction model, it is necessary to consider the reliability and the average width of the prediction section and the accuracy of the load average value, and therefore, the following four indexes are selected as the evaluation indexes of the model:
(1) prediction of gap coverage (PICP)
Wherein L istAnd UtRepresenting the upper and lower limits of the prediction interval, the larger the value of the PICP, the more the actual load values falling within the prediction interval.
(2) Prediction Interval Normalized Average Width (PINAW)
Wherein R represents the average width of the prediction interval. Generally, the smaller the value of PINAW is on the premise of ensuring that the PICP is as large as possible, the better the prediction accuracy is.
(3) Combined index (Coverage width-based criterion, CWC)
In order to meet the requirements of high coverage rate and narrow interval of the two indexes, the two indexes are converted into a comprehensive index for evaluation, and the comprehensive index is defined as follows:
CWC=PINAW×(1+γ×e(-η×(PICP-μ))) (25)
whereinMu represents that the confidence probability is 100 (1-alpha)%, eta is a penalty coefficient and is used for punishing the PICP < mu so as to ensure that the optimization result meets the decision requirement, and the smaller the value is, the better the prediction effect of the interval is.
(4) Mean Absolute percent Error (Mean Absolute percent Error, MAPE)
WhereinAndrespectively are the measured value and the predicted mean value of the load, and N is the number of sample points. Generally, a smaller MAPE value indicates a higher prediction accuracy of the model.
Further, based on the above embodiment, as a more complete embodiment, the method may further perform, at the same time, updating the model parameters of the load probability interval prediction model by using a recursive least square method based on a forgetting factor, and performing performance evaluation on the load probability interval prediction model by selecting one or more of a prediction gap coverage rate, a prediction interval regularized average width, a comprehensive index, and an average absolute percentage error, where specific contents refer to the above, and are not described herein again.
The prediction system provided by the embodiment of the present application is introduced below, and the prediction system described below and the prediction method described above may be referred to correspondingly.
Referring to fig. 3, fig. 3 is a schematic structural diagram of a prediction system of power load characteristics according to an embodiment of the present application, where the prediction system includes:
the data selection module 100 is used for performing feature selection on historical load data and new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
a sample establishing module 200, configured to establish a training sample according to the load optimal feature set;
a gaussian model building module 300, configured to build a gaussian process regression model using the training samples; wherein, the samples in the training samples are input variable matrixes and corresponding target values;
a prediction model establishing module 400, configured to establish a load probability interval prediction model at a specific confidence level according to the gaussian process regression model;
and the prediction module 500 is configured to input the data in the load optimal feature set, and output a prediction result by using the load probability interval prediction model.
The present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps provided by the above-described embodiments. The storage medium may include: various media such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, which can 12 store program codes.
The application also provides a power load characteristic prediction terminal, which may include a memory and a processor, where the memory stores a computer program, and when the processor calls the computer program in the memory, the steps provided in the foregoing embodiments may be implemented. Of course, the power load characteristic prediction terminal may further include various network interfaces, power supplies, and other components.
The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system provided by the embodiment, the description is relatively simple because the system corresponds to the method provided by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present application are explained herein using specific examples, which are provided only to help understand the method and the core idea of the present application. It should be noted that, for those skilled in the art, it is possible to make several improvements and modifications to the present application without departing from the principle of the present application, and such improvements and modifications also fall within the scope of the claims of the present application.
It is further noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Claims (8)
1. A method for predicting a characteristic of an electrical load, comprising:
performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
establishing a training sample according to the load optimal feature set; wherein, the samples in the training samples are input variable matrixes and corresponding target values;
establishing a Gaussian process regression model by using the training samples;
according to the Gaussian process regression model, a load probability interval prediction model under a specific confidence level is built;
inputting data in the load optimal characteristic set, and outputting a prediction result by using the load probability interval prediction model;
the method for selecting the characteristics of the historical load data and the new load data and outputting the load optimal characteristic set by using the K-Means characteristic extraction method comprises the following steps:
constructing a candidate characteristic set for the historical load data by using a K-Means characteristic extraction method;
performing feature clustering on the candidate feature set to form a first candidate feature set;
adding new load data into the first candidate feature set to form a second candidate feature set;
performing online feature classification and feature selection on the second candidate feature set, and outputting a load optimal feature set;
outputting a prediction result by using the load probability interval prediction model, and further comprising:
and updating the model parameters of the load probability interval prediction model by using a recursive least square method based on forgetting factors.
2. The prediction method according to claim 1, further comprising:
updating the load-optimized feature set each time the new load data is added to the first candidate feature set.
3. The prediction method of claim 1, wherein after building the training sample according to the load-optimized feature set, further comprising:
and updating the training samples by using a sliding time window technology, and ensuring that the number of the training samples is kept unchanged during updating.
4. The prediction method according to claim 1, wherein constructing a load probability interval prediction model at a certain confidence level according to the gaussian process regression model comprises:
establishing a relational expression between an input matrix and an output matrix according to the Gaussian process regression model;
establishing a finite set conforming to joint Gaussian distribution according to the relational expression;
determining a kernel matrix in the finite set, determining a distribution function of the target values;
determining the edge distribution of the distribution function, determining a hyper-parameter set according to the edge distribution, and obtaining a log-likelihood function of the Gaussian process regression model;
calculating the joint distribution of the input samples and the input variable matrix;
establishing a prediction model of the input sample according to the joint distribution;
and constructing a load probability interval prediction model under a specific confidence level according to a preset confidence level and the prediction model.
5. The prediction method according to any one of claims 1 to 4, further comprising:
and selecting one or more of the predicted clearance coverage rate, the normalized average width of the predicted interval, the comprehensive index and the average absolute percentage error to carry out performance evaluation on the load probability interval prediction model.
6. A power load characteristic prediction system capable of implementing a power load characteristic prediction method according to claim 1, comprising:
the data selection module is used for performing feature selection on the historical load data and the new load data by using a K-Means feature extraction method and outputting a load optimal feature set;
the sample establishing module is used for establishing a training sample according to the load optimal feature set; wherein, the samples in the training samples are input variable matrixes and corresponding target values;
the Gaussian model establishing module is used for establishing a Gaussian process regression model by using the training sample; the prediction model establishing module is used for establishing a load probability interval prediction model under a specific confidence level according to the Gaussian process regression model;
and the prediction module is used for inputting the data in the load optimal characteristic set and outputting a prediction result by using the load probability interval prediction model.
7. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the prediction method according to any one of claims 1 to 5.
8. A power load characteristic prediction terminal, characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the prediction method according to any one of claims 1 to 5 when calling the computer program in the memory.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810777121.5A CN108921358B (en) | 2018-07-16 | 2018-07-16 | Prediction method, prediction system and related device of power load characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810777121.5A CN108921358B (en) | 2018-07-16 | 2018-07-16 | Prediction method, prediction system and related device of power load characteristics |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108921358A CN108921358A (en) | 2018-11-30 |
CN108921358B true CN108921358B (en) | 2021-10-01 |
Family
ID=64411148
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810777121.5A Active CN108921358B (en) | 2018-07-16 | 2018-07-16 | Prediction method, prediction system and related device of power load characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108921358B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109945420B (en) * | 2019-03-26 | 2020-11-17 | 南京南瑞继保电气有限公司 | Air conditioner control method and device based on load prediction and computer storage medium |
CN112183565B (en) * | 2019-07-04 | 2023-07-14 | 创新先进技术有限公司 | Model training method, device and system |
CN111091242B (en) * | 2019-12-12 | 2021-10-08 | 浙江大学 | Optimal nonparametric interval prediction method for power load |
CN111898822B (en) * | 2020-07-30 | 2022-06-03 | 东北电力大学 | Charging load interval prediction method based on multi-correlation-day scene generation |
CN113052400A (en) * | 2021-04-23 | 2021-06-29 | 国网吉林省电力有限公司 | Short-term thermal load probability prediction method considering electrical-thermal coupling characteristics |
CN115130776B (en) * | 2022-07-13 | 2023-06-09 | 江南大学 | Wind power plant load prediction method based on Law asymmetric v-type TSVR |
CN117081082B (en) * | 2023-10-17 | 2024-01-23 | 国网上海市电力公司 | Active power distribution network operation situation sensing method and system based on Gaussian process regression |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855352A (en) * | 2012-08-17 | 2013-01-02 | 西北工业大学 | Method for clustering videos by using brain imaging space features and bottom layer vision features |
CN103974311A (en) * | 2014-05-21 | 2014-08-06 | 哈尔滨工业大学 | Condition monitoring data stream anomaly detection method based on improved gaussian process regression model |
CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
-
2018
- 2018-07-16 CN CN201810777121.5A patent/CN108921358B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102855352A (en) * | 2012-08-17 | 2013-01-02 | 西北工业大学 | Method for clustering videos by using brain imaging space features and bottom layer vision features |
CN103974311A (en) * | 2014-05-21 | 2014-08-06 | 哈尔滨工业大学 | Condition monitoring data stream anomaly detection method based on improved gaussian process regression model |
CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
Non-Patent Citations (6)
Title |
---|
An adaptive ensemble of on-line Extreme Learning Machines with variable forgetting factor for dynamic system prediction;Symone G.Soares 等;《Neurocomputing》;20160101;第171卷;693-707 * |
Estimation of the forgetting factor in kernel recursive least squares;Steven Van Vaerenbergh等;《2012 IEEE International Workshop on Machine Learning for Signal Processing》;20120926;1-6 * |
Online feature selection for high-dimensional class-imbalanced data;Peng Zhou等;《Knowledge-Based Systems》;20171115;第136卷;187-199 * |
基于K-均值聚类的无监督的特征选择方法;张莉 等;《计算机应用研究》;20050310(第3期);23-24,42 * |
基于在线学习高斯过程有杆泵油井动态液面建模方法研究;张栋;《中国优秀硕士学位论文全文数据库工程科技II辑》;20160815(第08期);B019-22 * |
基于改进高斯过程回归模型的短期负荷区间预测;宗文婷 等;《电力系统及其自动化学报》;20170815;第29卷(第8期);22-28 * |
Also Published As
Publication number | Publication date |
---|---|
CN108921358A (en) | 2018-11-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108921358B (en) | Prediction method, prediction system and related device of power load characteristics | |
CN110619420B (en) | Attention-GRU-based short-term residential load prediction method | |
CN112633316B (en) | Load prediction method and device based on boundary estimation theory | |
Porteiro et al. | Electricity demand forecasting in industrial and residential facilities using ensemble machine learning | |
JP2022033695A (en) | Method, device for generating model, electronic apparatus, storage medium and computer program product | |
CN110796307A (en) | Distributed load prediction method and system for comprehensive energy system | |
KR20200128232A (en) | Apparatus for predicting electricity demand and method thereof | |
CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
CN110796513A (en) | Multitask learning method and device, electronic equipment and storage medium | |
Porteiro et al. | Short term load forecasting of industrial electricity using machine learning | |
CN112990587A (en) | Method, system, equipment and medium for accurately predicting power consumption of transformer area | |
CN115099511A (en) | Photovoltaic power probability estimation method and system based on optimized copula | |
CN115271218A (en) | Carbon emission prediction method, device, equipment and medium based on electric carbon factor | |
CN112085256B (en) | Full-period load prediction method considering load jump | |
CN112651534A (en) | Method, device and storage medium for predicting resource supply chain demand | |
Gao et al. | A deep learning framework with spatial-temporal attention mechanism for cellular traffic prediction | |
Wang et al. | Three-dimensional maturity model of regional power users against the background of the ubiquitous power internet of things | |
CN114372835B (en) | Comprehensive energy service potential customer identification method, system and computer equipment | |
CN115204698A (en) | Real-time analysis method for power supply stability of low-voltage transformer area | |
CN114004530A (en) | Enterprise power credit score modeling method and system based on sequencing support vector machine | |
Li et al. | Decomposition-accumulation principle-based monthly electricity consumption forecasting approach using emd-xgboost hybrid model | |
Sun et al. | A new framework for short-term wind power probability forecasting considering spatial and temporal dependence of forecast errors | |
CN113837486B (en) | RNN-RBM-based distribution network feeder long-term load prediction method | |
CN116523148B (en) | Distribution network distribution transformer overload early warning method, device and equipment | |
Han et al. | Adaptive individual residential load forecasting based on deep learning and dynamic mirror descent |
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 |