CN116415715A - Prediction method for load clusters in short period of configuration - Google Patents
Prediction method for load clusters in short period of configuration Download PDFInfo
- Publication number
- CN116415715A CN116415715A CN202310228893.4A CN202310228893A CN116415715A CN 116415715 A CN116415715 A CN 116415715A CN 202310228893 A CN202310228893 A CN 202310228893A CN 116415715 A CN116415715 A CN 116415715A
- Authority
- CN
- China
- Prior art keywords
- load
- distribution transformer
- cluster
- short
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000013528 artificial neural network Methods 0.000 claims abstract description 39
- 230000000306 recurrent effect Effects 0.000 claims abstract description 35
- 238000003064 k means clustering Methods 0.000 claims abstract description 20
- 230000007787 long-term memory Effects 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 22
- 230000015654 memory Effects 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 8
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000013011 mating Effects 0.000 claims 1
- 230000006403 short-term memory Effects 0.000 abstract description 5
- 239000000203 mixture Substances 0.000 abstract description 3
- 230000008859 change Effects 0.000 abstract description 2
- 238000009472 formulation Methods 0.000 abstract description 2
- 239000010410 layer Substances 0.000 description 16
- 210000002569 neuron Anatomy 0.000 description 9
- 230000008569 process Effects 0.000 description 7
- 210000004027 cell Anatomy 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000003213 activating effect Effects 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 239000002356 single layer Substances 0.000 description 2
- 241001123248 Arma Species 0.000 description 1
- 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
- 230000006978 adaptation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a method for predicting a load cluster in a matched short period by acquiring pastTPreprocessing the original data of each distribution transformer load at each time point to obtain a sample data set; clustering the sample data of the distribution transformer in the sample data set by an improved K-means clustering method; based on a weighting coefficient method, the time sequence characteristics of each load curve are reduced in an equal proportion according to the distribution transformer capacity in each cluster, and various distribution transformer comprehensive training data are obtained; capturing the load characteristics of the same type of distribution transformer based on long-term and short-term memory recurrent neural network to obtain the respective prediction model of each type of distribution transformer, and distributingAnd inputting the comprehensive training data of the transformer as a data set into a prediction model to generate a load predicted value, and amplifying the predicted value in equal proportion to obtain a total load predicted value of the distribution transformer. The total load forecast value is used for accurately forecasting the load change of the power system in the short period, so that scientific data support is provided for the collaborative formulation of the power system dispatching and operation plans.
Description
Technical Field
The invention relates to the technical field of load prediction of power systems, in particular to a method for predicting a load cluster in a short period of time.
Background
Development of low-carbon clean energy power generation technology is urgent. However, along with the continuous improvement of clean energy permeability represented by wind/light, accurate load prediction is a basis for realizing safe and economic operation of a power system and scientific management of a power grid, is also an important guarantee for improving the utilization rate of power generation equipment and the effectiveness of economic dispatch, and has important significance for unit optimal combination, economic dispatch, optimal tide, electric power market transaction and the like.
The power system power load mainly comprises four kinds of power loads with different purposes, namely urban residential power load, commercial power load, rural power load and industrial power load, and factors influencing each load are different, for example: the industrial electric load is closely related to the characteristics of industry and the working mode, and rural electric loads are greatly influenced by seasons and climates. The power load is affected by various factors such as weather, economy, holidays and the like, has complex nonlinear characteristics, and if all the factors are used as input variables of a prediction model, the calculation complexity is increased, dimension disasters are caused, and the prediction accuracy is possibly reduced due to cross correlation among the variables, so that proper variables are selected as main study objects. The transformer clusters have extremely strong dispersibility, the climate and the load composition of the area where each transformer is located are different, and the influence of external factors on the transformer clusters is difficult to consider.
With the advent of artificial intelligence, the development of deep learning algorithms in data prediction applications has been on the go, and long and short memory recurrent neural networks (LSTM) represented by deep learning and neural networks have unique advantages in load prediction due to their good properties of capturing periodicity and time sequence. The LSTM can effectively process sequence data by using a long and short time sequence network, the time sequence of load data is reserved, but a large number of load characteristics are difficult to store by using the LSTM alone, and a single model is configured for each transformer, and because each model generates unavoidable errors in prediction; short-term load data can be predicted based on an ARMA model and an ARIMA model, short-sequence data can be effectively processed, the calculation speed is high, but the influence of the short-term load data on meteorological factors is not considered enough; moreover, the model has higher requirements on the stability of the original time sequence, and is not sensitive to external influence factors.
In view of this, there is a need for a method of matching short-term load cluster predictions.
Disclosure of Invention
The embodiment of the invention provides a distribution transformer short-term load cluster prediction method, which is used for at least solving the technical problem that the distribution transformer load prediction is not accurate enough due to nonlinear characteristics and transformer cluster dispersibility in the related technology.
According to an aspect of the embodiment of the invention, there is provided a method for predicting a matched short-term load cluster, including:
acquiring original data of each distribution transformer load at T time points in the past, and preprocessing the original data to obtain a sample data set;
clustering the sample data of the distribution transformer in the sample data set by an improved K-means clustering method;
based on a weighting coefficient method, the time sequence characteristics of each load curve are reduced in an equal proportion according to the distribution transformer capacity in each cluster, and various distribution transformer comprehensive training data are obtained;
based on the long-short-term memory recurrent neural network, capturing the load characteristics of the same type of distribution transformer to obtain a prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in equal proportion to obtain a total load predicted value of the distribution transformer.
Optionally, the original data is preprocessed to obtain a sample data set with 0 mean 1 variance.
Optionally, clustering the sample data of the distribution transformer in the sample data set by improving the K-means clustering method includes: the method comprises the steps of carrying out a first treatment on the surface of the
Step S21, randomly selecting one of the data points of the sample data set as a central point, and recording as u 1 ；
Step S22, calculating the Euclidean square distance between each data point x which is not selected in the sample data set and the selected center point;
step S23, randomly selecting a new data point from the sample data set as a new center u by using the weighted probability distribution n Wherein the probability of each selection point x is equal to the probability of the selection point x and the central point u n-1 The Euclidean square distance between the two is in direct proportion;
step S24, repeating steps S22-S23 until k centers are selected, k=n;
s25, calculating Euclidean square distances between the samples and each mean value vector according to the expression of the Euclidean square distances by taking the k selected centers as initial centers, and dividing the calculated Euclidean square distances into corresponding clusters according to the nearest mean value vector of the sample distances;
step S26, in combination with the family division in the step S25, calculating a new cluster average value by using the cluster average value to update the cluster average value vector, wherein the expression of the cluster average value is as follows:
wherein C is i Represents the i-th cluster, |C i I is the sum of the mean vectors of the ith cluster, x is the samples in the ith cluster,
the updated cluster average value;
and step S27, repeating the steps S25-S26 until the mean value vectors are all the latest, and obtaining k distribution transformer clusters.
Optionally, the expression of the euclidean square distance is:
wherein x is i And u is equal to ni Representing samples not yet selected and a center point, respectively, T being the sequence dimension, i.e., x i ＝[x 1 ,x 2 ...x T ]，u ni ＝[u n1 ,u n2 ...u nT ]，D 2 Representing the Euclidean squared distance between the data point and the center point.
Optionally, the expression of the comprehensive training data of the distribution transformer is:
wherein S is i For each capacity of the distribution transformer, L i For the load of each transformer,
representing the weight of the ith distribution transformer in the cluster, L train And (5) synthesizing training data for the distribution transformer of the model.
Optionally, capturing load characteristics of the same type of distribution transformer based on the long-term memory recurrent neural network, and obtaining a prediction model of each type of distribution transformer comprises the following steps:
building a long-term memory recurrent neural network;
the comprehensive training data of the distribution transformer is subjected to batch training treatment, the batch training treatment is input into a long-short-term memory recurrent neural network, and the average absolute error of the predicted value output by the long-short-term memory recurrent neural network and the real data set is calculated, namely, a Loss function is calculated to obtain a Loss value Loss VE ；
Judging Loss value Loss VE If the number of iterations is smaller than a certain fixed value or exceeds a limit range, entering the next step; if not, the error is reversely transmitted through the optimizer, the network parameters are updated, and the iteration times are added by one;
loss value Loss VE And (3) saving network grid parameters meeting requirements, and establishing a load prediction model of a plurality of long-short-period memory recurrent neural networks based on comprehensive training data of various distribution transformers according to the network grid parameters.
Optionally, loss value Loss VE The expression of (2) is:
wherein y is i 、
Respectively a predicted value and an actual value; and i εn, n is the sample set; m is the sequence dimension.
According to another aspect of the embodiment of the present invention, there is also provided a system for predicting a matched short-term load cluster, including:
the sample processing module is used for acquiring the original data of each distribution transformer load at the past T time points, and preprocessing the original data to obtain a sample data set;
the improved K-means clustering module is used for clustering the distributed sample data in the sample data set by an improved K-means clustering method;
the data processing module is used for reducing the time sequence characteristic of each load curve in equal proportion according to the distribution transformer capacity in each cluster based on a weighting coefficient method to obtain comprehensive training data of various distribution transformers;
and the prediction model module is used for capturing the load characteristics of the same type of distribution transformer based on the long-period memory recurrent neural network to obtain a respective prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in an equal proportion to obtain a total load predicted value of the distribution transformer.
According to another aspect of the embodiment of the present invention, there is further provided a computer readable storage medium, where the computer readable storage medium includes a stored program, and when the program runs, the device where the computer readable storage medium is controlled to execute the method for predicting a configuration short-term load cluster according to any one of the above.
According to another aspect of the embodiment of the present invention, there is further provided a processor, where the processor is configured to run a program, where the program executes any one of the foregoing methods for predicting a matched short-term load cluster.
Compared with the prior art, the invention has the following beneficial effects:
in the embodiment of the invention, the method for predicting the distribution and short-term load cluster obtains the original data of each distribution and transformation load at the past T time points, and preprocesses the original data to obtain a sample data set; clustering the sample data of the distribution transformer in the sample data set by an improved K-means clustering method; based on a weighting coefficient method, the time sequence characteristics of each load curve are reduced in an equal proportion according to the distribution transformer capacity in each cluster, and various distribution transformer comprehensive training data are obtained; based on the long-short-term memory recurrent neural network, capturing the load characteristics of the same type of distribution transformer to obtain a prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in equal proportion to obtain a total load predicted value of the distribution transformer. The total load forecast value is used for accurately forecasting the load change of the power system in the short period, so that scientific data support is provided for the collaborative formulation of the power system dispatching and operation plans.
The evaluation method can lead the reasonable development of the Internet of things technology in the power system, can solve the problem of load prediction of the distribution transformer caused by nonlinear characteristics and transformer cluster dispersibility, and has more guiding significance for the planning of power grid dispatching and maintenance planning by carrying out load prediction on the distribution transformer side by taking day as a time scale through the artificial intelligent prediction technology based on the neural network.
According to the invention, the problems of increased computational complexity and the like when the strong dispersion scene of the transformer multi-cluster are technically considered, and the weak clustering is avoided by adopting the traditional K-means clustering method and combining with the adjustment of the actual application scene to improve the K-means clustering method, so that the method has a certain reference value for the long sequence scene fitting.
The method is suitable for short-term load prediction, and by constructing the long-term memory recurrent neural network, the characteristics of multifactor, complexity, strong randomness and the like in the system are fully considered, the utilization rate of power generation equipment and the effectiveness of economic dispatching are improved, and a foundation is laid for realizing safe and economic operation of a power system and scientific management of a power grid.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawing in the description below is only one embodiment of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for matching short-lived load cluster predictions in accordance with an embodiment of the invention;
FIG. 2 is a schematic diagram of a discard regularization operation in accordance with an embodiment of the invention;
FIG. 3 is a block diagram of a single layer long and short term memory recurrent neural network according to an embodiment of the invention;
FIG. 4 is a flowchart of a long and short term memory recurrent neural network training process according to an embodiment of the invention;
FIG. 5 is a graph of various feature cluster profiles based on an improved K-means method in accordance with an embodiment of the invention;
fig. 6 is a diagram of a prediction effect of a distribution transformer cluster according to an embodiment of the present invention.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the present application described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
In accordance with an embodiment of the present invention, there is provided an embodiment of a method for adapting short-lived load cluster predictions, it being noted that the steps shown in the flowchart of the figures may be performed in a computer system, such as a set of computer-executable instructions, and, although a logical sequence is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than what is shown or described herein.
FIG. 1 is a flowchart of a method for predicting a matched short-term load cluster according to an embodiment of the invention, as shown in FIG. 1, the method includes the following steps:
step S1, acquiring original data of each distribution transformer load at T time points in the past, and preprocessing the original data to obtain a sample data set.
As an alternative embodiment, the raw data is subjected to preprocessing such as data cleaning, centering, standardization and the like, so as to obtain a sample data set with 0 mean value and 1 variance, and the normalization method is shown in formula 1.
In the method, in the process of the invention,
for normalized load data values, X t X is the original load data value min And X is max The minimum and maximum values of the original load dataset, respectively.
And S2, clustering the sample data of the distribution transformer by using an improved K-means clustering method based on the sample data of the distribution transformer preprocessed in the step S1.
As an alternative embodiment, step S2 specifically includes the following steps:
step S21, randomly selecting one of the data points of the sample data set preprocessed in the step S1 as a center point, and recording as u 1 ；
Step S22, calculating the Euclidean square distance between each data point x which is not selected in the sample data set and the selected central point, wherein the Euclidean square distance is shown in a formula (2).
Wherein x is i And u is equal to ni Representing samples not yet selected and a center point, respectively, T being the sequence dimension, i.e., x i ＝[x 1 ,x 2 ...x T ]，u ni ＝[u n1 ,u n2 ...u nT ]，D 2 Representing the Euclidean squared distance between the data point and the center point.
Step S23, randomly selecting a new data point from the sample data set processed in step S1 as a new center u by using the weighted probability distribution n Wherein each data point x is selected with a probability and a central point u n-1 The Euclidean square distance between the two is in direct proportion, and the expression is shown in the following formula.
Wherein P (u) i =x) represents the probability that data point X is selected, X represents the sample data set, u i Representing the newly selected center point of the present wheel, D (x) 2 Representing data points x and u i-1 The euclidean squared distance between them.
Step S24, repeating the steps S22 and S23 until k centers are selected;
s25, calculating Euclidean square distances between the samples and each mean value vector according to a formula (2) by taking the selected k centers as initial centers, and dividing the calculated Euclidean square distances into corresponding clusters according to the mean value vector with the nearest sample distance;
and S26, calculating a new cluster average value by using the cluster average value in combination with the cluster division of the step S25, and updating the cluster average value vector, wherein the formula is shown in a formula (4).
Wherein C is i Represents the i-th cluster, |C i I is the sum of the mean vectors of the ith cluster, x is the samples in the ith cluster,
and (5) the updated cluster average value.
And step S27, repeating the steps S25 and S26 until the mean value vectors are all the latest, and obtaining k distribution transformer clusters.
Step S3, based on step S2, the improved K-means clustering classification result and the weighting coefficient method are obtained, and the load curves { L } of the same class of distribution transformers are used 1 ,L 2 ,L 3 ,...,L m And performing equal-scale scaling according to the corresponding capacity to obtain comprehensive training data of the distribution transformer of each cluster, wherein the method is shown in a formula (5).
Wherein S is i For each capacity of the distribution transformer, L i For the load of each transformer,
representing the weight of the ith distribution transformer in the cluster, L train And (5) synthesizing training data for the distribution transformer of the model.
And S4, capturing load characteristics of the same type of distribution transformer based on the long-period memory recurrent neural network to obtain a prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in an equal proportion to obtain a total load predicted value of the distribution transformer.
As an alternative embodiment, as shown in fig. 4, the main steps of step S4 are as follows:
step S41, building a long-term memory recurrent neural network, setting a long-term memory recurrent neural network layer, a fully-connected neural network layer, discarding a regularization layer, setting an activation function, importing an optimizer and initializing grid parameters;
the long-term memory recurrent neural network layer is as follows:
wherein x is t To input the sample value of the layer, t is the sample index value, f t Representing a forgetting gate element vector, i t Represents an input gate cell vector, o t Representing the output gate cell vector, C t The cell state element vector is represented as,
is a cell state transition state element vector, h t For the hidden layer cell vector, w and b are the connection weight and bias between neurons in each cell respectively; the structure of the single-layer long-short-term memory recurrent neural network is shown in figure 3.
The fully connected neural network layer is as follows:
wherein n represents the number of characteristic input neurons, q represents the number of output neurons, p represents the number of hidden layer neurons, k represents a sample index value, w and b are respectively the connection weight and bias between neurons, f (·) represents an activation function expression, hi h (k) And ho h (k) Respectively representing the output of the h hidden layer before the activation function and the output after the activation function for the kth sample, yi o(k) And yo o (k) Respectively representing the output of the h output layer neuron before activating the function and the output after activating the function for the kth sample;
as shown in fig. 2, the discard regularization layer is as follows:
wherein r is i (l) Is an l-dimensional probability vector generated by Bernoulli function, the internal element value is 0 or 1, which represents the discarding and saving of the neuron, w and b are the connection weight and bias between the neurons, y (l) 、
Respectively representing the neuron input of the layer and the neuron input after random selection correction, ++>
And->
Respectively representing the output of the layer of neurons before activating the function and the output after activating the function;
the activation function layers based on ReLU and LeakyReLU are shown in the formulas (9) and (10) respectively:
the optimizer Adam of the long-term memory recurrent neural network is:
in the method, in the process of the invention,
weights and biases of network parameters at the nth iteration are respectively represented, +.>
And->
Second order gradient momentums accumulated in the previous n iterations of the loss function, beta 1 And beta 2 For the momentum gradient coefficient, alpha is the learning rate,
for the gradient correction amount, ε is a small constant to prevent the denominator from being 0.
Step S42, carrying out batch training treatment on the comprehensive training data of the distribution transformer, inputting the comprehensive training data of the distribution transformer into the long-short-term memory recurrent neural network obtained in the step S1, and calculating the average absolute error of the predicted value output by the long-short-term memory recurrent neural network and the real data set as a Loss function Loss VE Training, loss value Loss VE The formula of (2) is shown below.
Wherein y is i 、
Respectively a predicted value and an actual value, and i epsilon n, (n is a sample set), and m is a sequence dimension.
Step S43, judging Loss value Loss VE If the number of iterations is smaller than a certain fixed value or exceeds a limit range, entering the next step; if not, the error is reversely transmitted through the optimizer, the network parameters are updated, and the iteration times are added by one;
step S44, loss value Loss VE Network grid parameters meeting requirements are stored, and a load prediction model of a plurality of long-short-period memory recurrent neural networks based on comprehensive training data of various distribution transformers is established according to the network grid parameters;
and S45, driving various prediction models to generate load predicted values based on various historical load data, and amplifying the predicted values according to the equal proportion of the number of the distribution transformer to obtain the total load value of the cluster distribution transformer, wherein the method is shown in a formula (13).
L sum ＝n×L predict (13)
Wherein L is sum For the total load value of the cluster distribution transformer, n is the number of the cluster distribution transformer, L predict Load predictors are generated for the predictive model.
Example 2
One embodiment of the method for predicting the matched short-term load cluster based on the improved K-means clustering and the long-term memory recurrent neural network is described in detail so as to enable a person skilled in the art to better understand the method, and the method specifically comprises the following steps:
1) And acquiring data of each distribution transformer load at the past T time points, and preprocessing the original data such as data cleaning, centering and standardization. Based on the data of the power load of a distribution transformer in a certain place in 2019 of Yangjiang city power grid in Guangdong province of China, lagrange interpolation is carried out on the data in the blank of the original data, bad data are automatically screened and removed, max-min normalization processing is carried out on a data set, and the original load data are 15 minutes as intervals.
2) According to the improved K-means clustering algorithm, the distribution changes are classified into K classes based on sample data characteristics. And clustering load curves of the distribution transformer based on an improved K-means clustering algorithm, wherein the second type of load in the clustering result is shown in figure 5.
3) And reducing the time sequence characteristic of each load curve according to the equal proportion of the distribution transformer capacity in each cluster based on a weighting coefficient method to obtain comprehensive training data of various distribution transformers. Aiming at the transformer loads of the same type, a weighting coefficient method is used for fusing time sequence characteristics of each transformer, and the processed data are used as training data to construct a prediction model.
4) Based on the long-term memory recurrent neural network, capturing load characteristics of the same type of distribution transformer to obtain respective prediction models of each type of distribution transformer, driving each type of prediction models to generate load prediction values by using each type of historical load data, and amplifying the prediction values according to the equal proportion of the number of the distribution transformers to obtain the total load prediction values of the cluster distribution transformer. After the original load data is processed by using a weighting coefficient method, the long-term and short-term memory recurrent neural network is optimized as training data, the next daily load is used as a label value in the training process, the long-term and short-term memory recurrent neural network is 2 layers, the dimension of a hidden layer is 128, the learning rate is set to 0.001, the dropout is set to 0.3, the mean square error is selected as a loss function, the Adam optimizer is used for optimizing network parameters, the average absolute percentage error (Mean Absolute Percentage Error, MAPE) and the root mean square error (Root Mean Square Error, RMSE) are used as evaluation standards to verify the prediction accuracy of the model, the prediction error is shown in table 1, and the prediction result is shown in fig. 6.
TABLE 1 prediction error
Example 3
According to another aspect of the embodiment of the present invention, there is also provided a system for predicting a matched short-period load cluster, where the matched short-period load cluster prediction system applies the above method for predicting a matched short-period load cluster, and the matched short-period load cluster prediction system includes:
the sample processing module is used for acquiring the original data of each distribution transformer load at the past T time points, and preprocessing the original data to obtain a sample data set;
the improved K-means clustering module is used for clustering the distributed sample data in the sample data set by an improved K-means clustering method;
the data processing module is used for reducing the time sequence characteristic of each load curve in equal proportion according to the distribution transformer capacity in each cluster based on a weighting coefficient method to obtain comprehensive training data of various distribution transformers;
and the prediction model module is used for capturing the load characteristics of the same type of distribution transformer based on the long-period memory recurrent neural network to obtain a respective prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in an equal proportion to obtain a total load predicted value of the distribution transformer.
The present invention is not limited to the above embodiments, but is to be accorded the widest scope consistent with the principles and other features of the present invention.
Example 4
According to another aspect of the embodiment of the present invention, there is further provided a computer readable storage medium, where the computer readable storage medium includes a stored program, and when the program runs, the device in which the computer readable storage medium is controlled to execute the method for predicting the configured short-term load cluster according to any one of the above.
Alternatively, in this embodiment, the above-mentioned computer readable storage medium may be located in any one of the computer terminals in the computer terminal group in the computer network or in any one of the mobile terminals in the mobile terminal group, and the above-mentioned computer readable storage medium includes a stored program.
Optionally, the computer readable storage medium is controlled to perform the following functions when the program is run: the method comprises the steps of taking original data of each distribution transformer load at T time points, and preprocessing the original data to obtain a sample data set; clustering the sample data of the distribution transformer in the sample data set by an improved K-means clustering method; based on a weighting coefficient method, the time sequence characteristics of each load curve are reduced in an equal proportion according to the distribution transformer capacity in each cluster, and various distribution transformer comprehensive training data are obtained; based on the long-short-term memory recurrent neural network, capturing the load characteristics of the same type of distribution transformer to obtain a prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in equal proportion to obtain a total load predicted value of the distribution transformer.
Example 5
According to another aspect of the embodiment of the present invention, there is further provided a processor, configured to run a program, where the program executes the method for predicting a configured short-term load cluster according to any one of the above.
The embodiment of the invention provides equipment, which comprises a processor, a memory and a program stored in the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the method for predicting the configuration short-term load cluster.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology content may be implemented in other manners. The system embodiments described above are merely exemplary, and for example, the division of the units may be a logic function division, and there may be another division manner when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces and the indirect coupling or communication connection of units or modules may be in electrical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-0nlyMemory (ROM), a random access memory (RAM, randomAccessMemory), a removable hard disk, a magnetic disk, or an optical disk, or the like, which can store program codes.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (10)
1. The method for predicting the matched short-term load cluster is characterized by comprising the following steps of:
acquiring original data of each distribution transformer load at T time points in the past, and preprocessing the original data to obtain a sample data set;
clustering the sample data of the distribution transformer in the sample data set by an improved K-means clustering method;
based on a weighting coefficient method, the time sequence characteristics of each load curve are reduced in an equal proportion according to the distribution transformer capacity in each cluster, and various distribution transformer comprehensive training data are obtained;
based on the long-short-term memory recurrent neural network, capturing the load characteristics of the same type of distribution transformer to obtain a prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in equal proportion to obtain a total load predicted value of the distribution transformer.
2. The method for predicting the matched short-term load cluster according to claim 1, wherein a sample data set with 0 mean value and 1 variance is obtained after preprocessing the original data.
3. The distribution short-term load cluster prediction method according to claim 1, wherein clustering distribution-changed sample data in the sample data set by improving a K-means clustering method comprises: the method comprises the steps of carrying out a first treatment on the surface of the
Step S21, randomly selecting one of the data points of the sample data set as a central point, and recording as u 1 ；
Step S22, calculating the Euclidean square distance between each data point x which is not selected in the sample data set and the selected center point;
step S23, randomly selecting a new data point from the sample data set as a new center u by using the weighted probability distribution n Wherein the probability of each selection point x is equal to the probability of the selection point x and the central point u n-1 The Euclidean square distance between the two is in direct proportion;
step S24, repeating steps S22-S23 until k centers are selected, k=n;
s25, calculating Euclidean square distances between the samples and each mean value vector according to the expression of the Euclidean square distances by taking the k selected centers as initial centers, and dividing the calculated Euclidean square distances into corresponding clusters according to the nearest mean value vector of the sample distances;
step S26, in combination with the family division in the step S25, calculating a new cluster average value by using the cluster average value to update the cluster average value vector, wherein the expression of the cluster average value is as follows:
wherein C is i Represents the i-th cluster, |C i I is the sum of the mean vectors of the ith cluster, x is the samples in the ith cluster,
the updated cluster average value;
and step S27, repeating the steps S25-S26 until the mean value vectors are all the latest, and obtaining k distribution transformer clusters.
4. The configuration short-term load cluster prediction method according to claim 3, wherein the expression of the euclidean squared distance is:
wherein x is i And u is equal to ni Representing samples not yet selected and a center point, respectively, T being the sequence dimension, i.e., x i ＝[x 1 ,x 2 ...x T ]，u ni ＝[u n1 ,u n2 ...u nT ]，D 2 Representing the Euclidean squared distance between the data point and the center point.
5. The method for predicting the distribution short-term load cluster according to claim 1, wherein the expression of the comprehensive training data of the distribution transformer is:
wherein S is i For each capacity of the distribution transformer, L i For the load of each transformer,
representing the weight of the ith distribution transformer in the cluster, L train And (5) synthesizing training data for the distribution transformer of the model.
6. The method for predicting the distribution and short-term load cluster according to claim 1, wherein capturing the load characteristics of the same class of distribution and transformation based on the long-term memory recurrent neural network, and obtaining the respective prediction model of each class of distribution and transformation comprises:
building a long-term memory recurrent neural network;
the comprehensive training data of the distribution transformer is subjected to batch training treatment, the batch training treatment is input into a long-short-term memory recurrent neural network, and the average absolute error of the predicted value output by the long-short-term memory recurrent neural network and the real data set is calculated, namely, a Loss function is calculated to obtain a Loss value Loss VE ；
Judging Loss value Loss VE If the number of iterations is smaller than a certain fixed value or exceeds a limit range, entering the next step; if not, the error is reversely transmitted through the optimizer, the network parameters are updated, and the iteration times are added by one;
loss value Loss VE And (3) saving network grid parameters meeting requirements, and establishing a load prediction model of a plurality of long-short-period memory recurrent neural networks based on comprehensive training data of various distribution transformers according to the network grid parameters.
7. The method for predicting a short-lived load cluster of claim 6 wherein the Loss value Loss VE The expression of (2) is:
wherein y is i 、
Respectively a predicted value and an actual value; and i εn, n is the sample set; m is the sequence dimension.
8. A mating short-term load cluster prediction system, comprising:
the sample processing module is used for acquiring the original data of each distribution transformer load at the past T time points, and preprocessing the original data to obtain a sample data set;
the improved K-means clustering module is used for clustering the distributed sample data in the sample data set by an improved K-means clustering method;
the data processing module is used for reducing the time sequence characteristic of each load curve in equal proportion according to the distribution transformer capacity in each cluster based on a weighting coefficient method to obtain comprehensive training data of various distribution transformers;
and the prediction model module is used for capturing the load characteristics of the same type of distribution transformer based on the long-period memory recurrent neural network to obtain a respective prediction model of each type of distribution transformer, inputting the comprehensive training data of the distribution transformer as a data set into the prediction model to generate a load predicted value, and amplifying the predicted value in an equal proportion to obtain a total load predicted value of the distribution transformer.
9. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored program, wherein the program, when run, controls a device in which the computer-readable storage medium is located to perform the method for matching short-term load cluster prediction according to any one of claims 1 to 7.
10. A processor, characterized in that the processor is configured to run a program, wherein the program runs to perform the method for predicting a matched short-term load cluster according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310228893.4A CN116415715A (en) | 2023-03-10 | 2023-03-10 | Prediction method for load clusters in short period of configuration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310228893.4A CN116415715A (en) | 2023-03-10 | 2023-03-10 | Prediction method for load clusters in short period of configuration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116415715A true CN116415715A (en) | 2023-07-11 |
Family
ID=87055694
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310228893.4A Pending CN116415715A (en) | 2023-03-10 | 2023-03-10 | Prediction method for load clusters in short period of configuration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116415715A (en) |
-
2023
- 2023-03-10 CN CN202310228893.4A patent/CN116415715A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tang et al. | Short‐term power load forecasting based on multi‐layer bidirectional recurrent neural network | |
| Ke et al. | Short-term electrical load forecasting method based on stacked auto-encoding and GRU neural network | |
| CN110263995B (en) | Distribution transformer overload prediction method considering load increase rate and user power utilization characteristics | |
| CN109886492A (en) | Photovoltaic power generation power prediction model and its construction method based on Attention LSTM | |
| CN108280551A (en) | A kind of photovoltaic power generation power prediction method using shot and long term memory network | |
| CN110490385A (en) | The unified prediction of electric load and thermic load in a kind of integrated energy system | |
| CN110674999A (en) | Cell load prediction method based on improved clustering and long-short term memory deep learning | |
| Kolhe et al. | GA-ANN for short-term wind energy prediction | |
| Chen et al. | MultiCycleNet: multiple cycles self-boosted neural network for short-term electric household load forecasting | |
| CN115222106A (en) | User day-ahead load prediction method of self-adaptive model | |
| CN109190786A (en) | A kind of combination forecasting method of photovoltaic efficiency | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN117151770A (en) | Attention mechanism-based LSTM carbon price prediction method and system | |
| Xu et al. | Short-term electricity consumption forecasting method for residential users based on cluster classification and backpropagation neural network | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| Zuo | Integrated forecasting models based on LSTM and TCN for short-term electricity load forecasting | |
| Lv et al. | Research on short-term load forecasting approach for smart grid | |
| CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
| CN116415715A (en) | Prediction method for load clusters in short period of configuration | |
| CN111680818B (en) | Short-term reactive load prediction method and system | |
| CN113128754A (en) | GRU neural network-based residential water use prediction system and prediction method | |
| Wang et al. | Short term load forecasting: A dynamic neural network based genetic algorithm optimization | |
| Liu et al. | Study on electricity market price forecasting with large-scale wind power based on LSTM | |
| CN111915084A (en) | Hybrid photovoltaic power generation power prediction method and system based on neural network |
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 |