CN112884077A - Garden short-term load prediction method based on dynamic time regression clustering of shapes - Google Patents
Garden short-term load prediction method based on dynamic time regression clustering of shapes Download PDFInfo
- Publication number
- CN112884077A CN112884077A CN202110323783.7A CN202110323783A CN112884077A CN 112884077 A CN112884077 A CN 112884077A CN 202110323783 A CN202110323783 A CN 202110323783A CN 112884077 A CN112884077 A CN 112884077A
- Authority
- CN
- China
- Prior art keywords
- data
- load
- short
- clustering
- park
- 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 35
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000006399 behavior Effects 0.000 claims abstract description 17
- 238000012795 verification Methods 0.000 claims abstract description 14
- 238000005265 energy consumption Methods 0.000 claims abstract description 11
- 230000006403 short-term memory Effects 0.000 claims abstract description 11
- 230000007787 long-term memory Effects 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000005611 electricity Effects 0.000 claims description 22
- 230000007704 transition Effects 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 13
- 230000015654 memory Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 108091026890 Coding region Proteins 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 3
- 238000007621 cluster analysis Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013277 forecasting method Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000010200 validation analysis Methods 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
- G06F18/21—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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a garden short-term load prediction method based on dynamic time regression clustering of shapes, which comprises the following steps: acquiring and preprocessing data; finishing the cluster analysis of the main energy consumption behaviors of different units in the park; predicting the load behavior clustering result of the users in different parks at the next moment; obtaining a final short-term load prediction result of the park; dividing the load data and the corresponding data into a training set and a verification set; taking the actual load of the input vector at a fixed time as a training target of model output; selecting the model with the best performance in the verification set as a training result; and (4) running the long-term and short-term memory model in the actual environment, and adding the latest data into the training set to train the model again when the predicted value and the actual value have larger deviation. Has the advantages that: the method provided by the invention effectively integrates the daily energy consumption behavior characteristics of the park users to predict the short-term load, and can effectively improve the prediction precision.
Description
Technical Field
The invention relates to the field of short-term load prediction, in particular to a park short-term load prediction method based on dynamic time regression clustering of shapes.
Background
With the continuous advance of new power system innovation, incremental distribution business innovation becomes the most concerned topic of governments and society. In recent two years, more incremental distribution network trial projects are available, and social high-quality capital participation is continuously attracted. Most of the pilot project owners are composed of high and new industry parks, circular economy parks and industrial parks. The main body electrical loads are relatively concentrated and have strong regularity, so that the characteristics of the electrical loads of the main bodies in the pilot project have very important significance in the current incremental distribution network market construction.
In the incremental power supply market competition, the power supply economy and reliability are decisive factors. In the aspect of reliability, power supply reliability of different power supply modes is different, and cost is different, but under the current policy, the function of the power grid cost for improving reliability cannot be embodied, so that the competition depending on the power supply reliability is relatively difficult to grasp before the current electricity price market mechanism is not completely mature; in the aspect of economy, the power supply requirement of a park is met in a targeted manner by accurate measures, and the increase of the input-output ratio of an incremental distribution network is an important influence factor in the market at the present stage, so that the load prediction of the incremental distribution network is very important. Aiming at the traditional forecasting method of the garden, a load forecasting method based on machine learning and deep learning is mostly adopted. The technologies usually accept longer load sequences, and a complex learning mechanism is utilized to extract the nonlinear relation in the sequences so as to form a better load prediction result. However, in the load prediction of the park, because different unit bodies in the park have strong energy consumption relativity, if the power consumption behaviors of different units are subjected to cluster analysis and the results are integrated into the load prediction process, the whole load prediction level of the park can be greatly improved.
An effective solution to the problems in the related art has not been proposed yet.
Disclosure of Invention
Aiming at the problems in the related art, the invention provides a garden area short-term load prediction method based on shape dynamic time integral clustering, so as to overcome the technical problems in the prior related art.
Therefore, the invention adopts the following specific technical scheme:
the method for forecasting the park short-term load based on the dynamic time regression clustering of the shape comprises the following steps:
s1, acquiring load data and corresponding data of the power system, and preprocessing the load data and the corresponding data;
s2, constructing a dynamic time normalization clustering method based on the multi-scale information fusion convolutional neural network model, and completing clustering analysis of main body energy consumption behaviors of different units in the park;
s3, predicting the next-time load behavior clustering result of users in different parks by using a hidden Markov model;
s4, merging the clustering results into a long-short term memory model, performing short term load prediction on different users in the park through the long-short term memory model, and adding the prediction results of the different users to obtain the final short term load prediction result of the park;
s5, dividing the load data and the corresponding data into a training set and a verification set;
s6, selecting a training set vector sequence with a fixed length as long-short term memory model input, taking the actual load of the input vector in a fixed time as a training target of the model output, and converging the model through multiple iterative training;
s7, verifying the trained long and short term memory model by using a verification set, adjusting model parameters by comparing the precision and the error of the test set and the verification set, and selecting the model with the best performance of the verification set as a training result by training for multiple times;
s8, operating the long-term and short-term memory model in the actual environment, and adding the latest data into the training set to train the model again when the predicted value and the actual value have larger deviation;
the corresponding data in S1 includes, but is not limited to, climate data and holiday data.
Further, the step of acquiring the load data and the corresponding data of the power system in S1, and preprocessing the load data and the corresponding data thereof includes the following steps:
s11, acquiring load data of the power system and corresponding data of the load data;
s12, converting time data in the load data and the corresponding data into numerical data, standardizing all the data, performing mean interpolation on missing values in the data, and converting holiday data into binary data;
s13, forming an input vector sequence according to the time sequence;
s14, the system load in the future period of time of the preset time is used as the prediction target.
Further, the formula of the multi-scale information fusion convolutional neural network model in S2 is as follows: p is a radical ofi＝(ni,mi)∈[1:N]×[1:N]，1<＝i<＝L。
Further, constructing a dynamic time normalization clustering method in S2, and completing cluster analysis of different unit subject energy consumption behaviors of the park includes the following steps:
s21, setting a boundary condition, wherein,
p1＝(1,1)，
and p isL＝(N,N)；
S22, each point on the prescribed path is monotonous to go with time, and niAnd miSatisfies n1≤ni≤nLAnd m1≤mi≤mL；
S23, defining a point p on the pathi＝(ni,mi)，
The next point pi+1＝(ni+1,mi+1) Satisfies ni+1-niM is less than or equal to 1i+1-mi≤1；
S24, calculating the total cost function of the regular path between the time series X and Y, wherein the formula of the function is
S25, calculating the distance between every two points in the time sequence X, Y, storing by using a cost matrix C, and calculating the total cost of the optimal regular path through recursion, wherein the recursion formula is as follows:
further, the step of calculating the total cost of the optimal warping path in S25 includes the following steps:
setting a user load curve of a park as X, the number of clusters as K, and setting a prototype of each cluster as muk(ii) a The clustering algorithm based on DTW finds that the sum of DTW distances of K clusters is the lowest, namely:
further, the step of predicting the load behavior clustering result of the user in different parks at the next moment by using the hidden markov model in the step S3 includes calculating a state transition matrix of n steps of the markov chain and predicting the state of the next step;
wherein, the state transition matrix of n steps of the Markov chain is calculated by the following steps:
selecting historical codes as the basis for dividing the system state, and setting that the code sequence contains r states in total and recording as s1,s2,…,sr；
When the number of clusters is K, a certain state siHas a value range of [0, K-1 ]]Coded data of history from siThe state is transferred to s through n stepsjProbability of state Pij(n) is: pij(n)＝Mij(n)/MiWherein M isij(n) is the coding sequence consisting of siThe state is transferred to s through n stepsjNumber of states, MiRepresenting the coding sequence in siTotal number of states;
based on the above formula, a state transition matrix p (n) of n steps of the markov chain is obtained as:
Further, the predicting the state of the next step comprises the following steps:
in the state s of step rrBased on this, and obtains the state s from the state transition matrix P (n)rMaximum probability of transfer, i.e. Max (P)rl(n)), where l ═ 1, 2, … r, finally slThe state is the state of r +1 step.
Further, the S4 middle and long short term memory model includes three parts:
a first part: clustering daily power utilization curves of users in the garden to obtain K clusters, and coding the clusters;
a second part: forecasting the electricity utilization curve codes of the park users by using a Markov chain model, and forecasting the electricity utilization codes of the r +1 th day by using the electricity utilization codes of the r days for each park user to obtain a prototype of the electricity utilization curve of the r +1 th day;
and a third part: the electricity utilization code predicted by the Markov chain is used as a characteristic, historical load data is combined, and a long-term and short-term memory network is utilized to predict the short-term load of the garden user.
The invention has the beneficial effects that: the invention provides a garden short-term load prediction method based on dynamic time regression clustering of shapes, which is used for clustering main daily electricity curves of different units of a garden and coding clustering results. And predicting the electricity utilization curve code of the future day by using a Markov chain model according to the code of the cluster to which the unit main body electricity utilization curve belongs in the past ten days, so as to predict the electricity utilization curve prototype of the future day. The method is characterized in that the code is used, and the LSTM network is utilized to predict the future short-term load according to the historical electricity consumption data of different users in the park, so that the short-term load prediction of the whole situation of the park is realized. The method provided by the invention effectively integrates the daily energy consumption behavior characteristics of the park users to carry out short-term load prediction, so that the prediction precision can be effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow diagram of a campus short term load prediction method based on shape-based dynamic time-warping clustering, according to an embodiment of the present invention;
FIG. 2 is a diagram of a long-short term memory model network architecture;
fig. 3 is a schematic diagram of several commonly used neural network models.
Detailed Description
For further explanation of the various embodiments, the drawings which form a part of the disclosure and which are incorporated in and constitute a part of this specification, illustrate embodiments and, together with the description, serve to explain the principles of operation of the embodiments, and to enable others of ordinary skill in the art to understand the various embodiments and advantages of the invention, and, together with the description, reference is made to the figures and wherein the elements are not drawn to scale and wherein like reference numerals are generally used to designate like elements.
According to the embodiment of the invention, the garden short-term load prediction method based on the dynamic time regression clustering of the shape is provided, the advantages of various models are obtained by fusing the similarity characteristics of the energy consumption behaviors of different users, and the garden short-term load prediction precision is effectively improved.
The present invention will now be further described with reference to the accompanying drawings and specific embodiments, wherein as shown in fig. 1 to 3, a method for forecasting a short-term load of a campus based on dynamic time-rounded clustering of shapes according to an embodiment of the present invention comprises the following steps:
step S1: acquiring load data and corresponding data of the power system, and preprocessing the load data and the corresponding data (data acquisition and preprocessing);
step S11: using the load data of the electricity network 2014 and 2018 somewhere, the accuracy is statistics per hour. Meanwhile, daily maximum air temperature and daily minimum air temperature of the region and holiday data are collected.
Step S12: time data is converted into numerical data, all data is subjected to normalization processing, missing values are subjected to mean interpolation processing, and holiday data is converted into binary data (preprocessing methods include, but are not limited to, missing value mean interpolation, data normalization, one-hot encoding, and time stamp conversion).
Step S13: forming an input vector sequence according to the time sequence, namely, the data of each hour is a vector and is arranged according to the time sequence;
step S14: the prediction target is the power system load at any time in the future of 6 hours.
Step S2: constructing a dynamic time normalization clustering method based on a multi-scale information fusion convolutional neural network model, and completing clustering analysis of energy consumption behaviors of different unit bodies in the park (constructing the multi-scale information fusion convolutional neural network model);
pi＝(ni,mi)∈[1:N]×[1:N](1<＝i<l). The regular path is not chosen at will and needs to satisfy the following conditions: (ii) a
Step S21: setting a boundary condition: wherein p is1Is (1,1), and pL＝(N,N)。
Step S22: each point on the prescribed path must be monotonic over time, so niAnd miSatisfies n1≤ni≤nL，m1≤mi≤mL；
Step S23: the specified continuity: for a point p on the pathi＝(ni,mi) Next point pi+1＝(ni+1,mi+1) Satisfies ni+1-n i1 or less and mi+1-mi≤1。
Step S24: calculating between time series X, YTotal cost function c of regular pathp(X, Y) is:
step S25: and solving by using a dynamic programming idea. Firstly, the distance between every two points in the time series X and Y is calculated and is stored by utilizing a cost matrix C. The total cost of the optimal regular path can be calculated recursively, and the recursive formula is as follows:
final DTW (x)N,yN) I.e. the total cost of the best warping path. Given a user load curve X of a park and the number of clusters K, the prototype of each cluster is set to be mukThe objective of the DTW-based clustering algorithm is to find that the sum of DTW distances (KC) of K clusters is minimal, i.e.:
therefore, the power consumption situation at each same moment in the power consumption curve in each cluster obtained by clustering the power consumption curve through a Dynamic Time Warping (DTW) algorithm may be different greatly, but the overall power consumption rules are similar.
Step S3: and predicting the load behavior clustering result of the users in different parks at the next moment by using a hidden Markov model.
The Markov chain process is based on the system state transition rule, researches and analyzes the development trend of the object, and accordingly deduces the most probable state of the object in the future.
Selecting historical codes as the basis for dividing the system state, and setting that the coded sequences contain r states which are marked as s1,s2,…,sr. When the number of clusters is K, a certain state siHas a value range of [0, K-1 ]]. Historical coded data from siThe state is transferred to s through n stepsjProbability of state Pij(n) is: pij(n)＝Mij(n)/Mi
Wherein M isij(n) is the coding sequence consisting of siThe state is transferred to s through n stepsjNumber of states, MiRepresenting the coding sequence in siThe total number of states.
Based on the above formula, a state transition matrix p (n) of n steps of the markov chain can be obtained as:
wherein: pij(n)≥0，
Calculating a transition matrix, predicting the state of the r +1 step, and calculating the state s of the r-th steprBased on finding the state s from the state transition matrixrMaximum probability of transition, i.e. Max (P)rl(n)), wherein l is 1, 2, … r. Final slThe state is then the most likely state for step r + 1.
Step S4: and merging the clustering results into a long-term and short-term memory model to predict the short-term load of different users in the park, and finally adding the prediction results of the different users to obtain the final short-term load prediction result of the park.
As shown in fig. 2, the overall model mainly comprises three parts. The first part is to cluster the daily power utilization curves of users in the garden to obtain K clusters and encode the clusters (0-K-1); predicting power utilization curve codes of park users by using a Markov chain model, predicting power utilization codes of the r +1 th day by using power utilization codes of r days for each park user, and aiming at obtaining a prototype of the power utilization curve of the r +1 th day; and the third part is to predict the short-term load of the park users, the electricity utilization code predicted by the Markov chain is used as a characteristic, and the long-term short-term memory network (LSTM) is used for predicting the short-term load by combining with historical load data.
Step S5: the data set is partitioned into two parts: the training set is used for training the model, and the verification set is used for verifying the training result. The first 80% of the data set was used as the training set and the last 20% as the validation set for this case.
Step S6: selecting a training set vector sequence with a fixed length of 240 as model input, taking an actual load 6 hours after the input vector as a training target of the model output, and training the model to be convergent through multiple iterations;
step S7: verifying the trained model by using a verification set, adjusting model parameters by comparing the precision and the error of the test set and the verification set, and selecting the model with the best performance of the verification set as a training result by training for multiple times;
step S8: and (4) running the model in an actual environment, and adding the latest data into the training set to train the model again when the predicted value and the actual value have larger deviation.
In order to verify the effectiveness of the method, several common neural network models are compared in the case, the result is shown in fig. 3, and the result shows that the prediction accuracy of the garden short-term load prediction method based on the dynamic time integral clustering of the shape is the highest in the several models, which shows that the method can effectively improve the accuracy of the short-term load prediction.
In summary, the invention provides a garden short-term load prediction method based on dynamic time normalization clustering of shapes, which clusters daily power curves of different unit bodies in a garden and encodes clustering results. And predicting the electricity utilization curve code of the future day by using a Markov chain model according to the code of the cluster to which the unit main body electricity utilization curve belongs in the past ten days, so as to predict the electricity utilization curve prototype of the future day. The method is characterized in that the code is used, and the LSTM network is utilized to predict the future short-term load according to the historical electricity utilization data of different users in the park, so that the short-term load prediction of the whole situation of the park is realized. The method provided by the invention effectively integrates the daily energy consumption behavior characteristics of the park users to predict the short-term load, so that the prediction precision can be effectively improved
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (8)
1. The campus short-term load prediction method based on the dynamic time regression clustering of the shape is characterized by comprising the following steps of:
s1, acquiring load data and corresponding data of the power system, and preprocessing the load data and the corresponding data;
s2, constructing a dynamic time normalization clustering method based on the multi-scale information fusion convolutional neural network model, and completing clustering analysis of main body energy consumption behaviors of different units in the park;
s3, predicting the load behavior clustering result of the users in different parks at the next moment by utilizing a hidden Markov model;
s4, merging the clustering results into a long-short term memory model, performing short term load prediction on different users in the park through the long-short term memory model, and adding the prediction results of the different users to obtain the final short term load prediction result of the park;
s5, dividing the load data and the corresponding data into a training set and a verification set;
s6, selecting a training set vector sequence with a fixed length as long-short term memory model input, taking the actual load of the input vector in a fixed time as a training target of the model output, and converging the model through multiple iterative training;
s7, verifying the trained long and short term memory model by using a verification set, adjusting model parameters by comparing the precision and the error of the test set and the verification set, and selecting the model with the best performance of the verification set as a training result by training for multiple times;
s8, operating the long-term and short-term memory model in the actual environment, and adding the latest data into the training set to train the model again when the predicted value and the actual value have large deviation;
the corresponding data in S1 includes, but is not limited to, climate data and holiday data.
2. The method for forecasting park short-term load based on shape dynamic time regression clustering according to claim 1, wherein the step of obtaining and pre-processing load data and corresponding data of the power system in S1 comprises the following steps:
s11, acquiring load data of the power system and corresponding data of the load data;
s12, converting time data in the load data and the corresponding data into numerical data, standardizing all the data, performing mean interpolation on missing values in the data, and converting holiday data into binary data;
s13, forming an input vector sequence according to the time sequence;
s14, the system load in the future period of time of the preset time is used as the prediction target.
3. The method for forecasting the campus short-term load of shape-based dynamic time regression clustering according to claim 1, wherein the formula of the multi-scale information fusion convolutional neural network model in S2 is as follows:
pi＝(ni,mi)∈[1:N]×[1:N]，1<＝i<＝L。
4. the method for forecasting the short-term load of the park based on the dynamic time regression clustering of the shape according to claim 3, wherein the dynamic time regression clustering method is constructed in S2, and the step of completing the clustering analysis of the main energy consumption behaviors of different units of the park comprises the following steps:
s21, setting a boundary condition, wherein,
p1＝(1,1)，
and p isL＝(N,N)；
S22, each point on the prescribed path is monotonous to go with time, and niAnd miSatisfies n1≤ni≤nLAnd m1≤mi≤mL；
S23, defining a point p on the pathi＝(ni,mi)，
The next point pi+1＝(ni+1,mi+1) Satisfies ni+1-niM is less than or equal to 1i+1-mi≤1；
S24, calculating the total cost function of the regular path between the time series X and Y, wherein the formula of the function is
S25, calculating the distance between every two points in the time sequence X, Y, and storing by using a cost matrix C, calculating the total cost of the optimal regular path through recursion, wherein the recursion formula is as follows:
5. the method for campus short term load prediction based on shape dynamic time regression clustering as claimed in claim 4, wherein said step of calculating total cost of optimal warping path in S25 is followed by the steps of:
setting a user load curve of a park as X, the number of clusters as K, and setting a prototype of each cluster as muk；
The clustering algorithm based on DTW finds that the sum of DTW distances of K clusters is the lowest, namely:
6. the campus short-term load prediction method based on shape-based dynamic time-rounded clustering of claim 1, wherein the predicting the load behavior clustering result at the next moment of different campus users by using hidden markov models in S3 includes calculating n-step state transition matrices of markov chains and predicting the state of the next step;
wherein, the state transition matrix of n steps of the Markov chain is calculated by the following steps:
selecting historical codes as the basis for dividing the system state, and setting that the code sequence contains r states in total and recording as s1,s2,…,sr；
When the number of clusters is K, a certain state siHas a value range of [0, K-1 ]]Historical coded data from siThe state is transferred to s through n stepsjProbability of state Pij(n) is: pij(n)＝Mij(n)/MiWherein M isij(n) is the coding sequence consisting of siThe state is transferred to s through n stepsjNumber of states, MiRepresenting the coding sequence in siTotal number of states;
based on the above formula, a state transition matrix p (n) of n steps of the markov chain is obtained as:
7. The method of shape-based dynamic time-rounded clustering campus short-term load prediction according to claim 6, wherein said predicting the status of the next step comprises the steps of:
in the state s of step rrBased on this, and obtains the state s from the state transition matrix P (n)rMaximum probability of transition, i.e. Max (P)rl(n)), where l ═ 1, 2, … r, finally slThe state is the state of r +1 step.
8. The method of claim 1, wherein the S4 medium-long short term memory model comprises three parts:
a first part: clustering daily power utilization curves of users in the garden to obtain K clusters, and coding the clusters;
a second part: forecasting the electricity utilization curve codes of the park users by using a Markov chain model, and forecasting the electricity utilization codes of the r +1 th day by using the electricity utilization codes of the r days for each park user to obtain a prototype of the electricity utilization curve of the r +1 th day;
and a third part: the electricity utilization code predicted by the Markov chain is used as a characteristic, historical load data is combined, and a long-term and short-term memory network is used for predicting the short-term load of the park users.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110323783.7A CN112884077A (en) | 2021-03-26 | 2021-03-26 | Garden short-term load prediction method based on dynamic time regression clustering of shapes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110323783.7A CN112884077A (en) | 2021-03-26 | 2021-03-26 | Garden short-term load prediction method based on dynamic time regression clustering of shapes |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112884077A true CN112884077A (en) | 2021-06-01 |
Family
ID=76042403
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110323783.7A Pending CN112884077A (en) | 2021-03-26 | 2021-03-26 | Garden short-term load prediction method based on dynamic time regression clustering of shapes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112884077A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114091338A (en) * | 2021-11-24 | 2022-02-25 | 国网江苏省电力有限公司泰州供电分公司 | Method and device for establishing power load decomposition model |
| CN115305976A (en) * | 2022-10-12 | 2022-11-08 | 铁科院(深圳)检测工程有限公司 | Intelligent loading system for static load experiment of large-diameter pile |
| CN117239746A (en) * | 2023-11-16 | 2023-12-15 | 国网湖北省电力有限公司武汉供电公司 | Power load prediction method and system based on machine learning |
-
2021
- 2021-03-26 CN CN202110323783.7A patent/CN112884077A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114091338A (en) * | 2021-11-24 | 2022-02-25 | 国网江苏省电力有限公司泰州供电分公司 | Method and device for establishing power load decomposition model |
| CN115305976A (en) * | 2022-10-12 | 2022-11-08 | 铁科院(深圳)检测工程有限公司 | Intelligent loading system for static load experiment of large-diameter pile |
| CN115305976B (en) * | 2022-10-12 | 2022-12-06 | 铁科院(深圳)检测工程有限公司 | Intelligent loading system for static load experiment of large-diameter pile |
| CN117239746A (en) * | 2023-11-16 | 2023-12-15 | 国网湖北省电力有限公司武汉供电公司 | Power load prediction method and system based on machine learning |
| CN117239746B (en) * | 2023-11-16 | 2024-01-30 | 国网湖北省电力有限公司武汉供电公司 | Power load prediction method and system based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109754113B (en) | Load prediction method based on dynamic time warping and long-and-short time memory | |
| CN112884077A (en) | Garden short-term load prediction method based on dynamic time regression clustering of shapes | |
| CN110969290B (en) | Runoff probability prediction method and system based on deep learning | |
| CN111160617B (en) | Power daily load prediction method and device | |
| CN109492748B (en) | Method for establishing medium-and-long-term load prediction model of power system based on convolutional neural network | |
| CN113344288B (en) | Cascade hydropower station group water level prediction method and device and computer readable storage medium | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN114511132A (en) | Photovoltaic output short-term prediction method and prediction system | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN115115125A (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
| CN115983710A (en) | High-proportion new energy access electric power system infrastructure project decision method and system | |
| CN115907131A (en) | Method and system for building electric heating load prediction model in northern area | |
| CN114219126B (en) | Small hydropower infiltration area network load supply prediction method based on residual error correction | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN117407681B (en) | Time sequence data prediction model establishment method based on vector clustering | |
| CN115018111A (en) | Wind power prediction method and system integrating deep learning and self-adaptive modeling mechanisms | |
| CN108694475A (en) | Short-term time scale photovoltaic cell capable of generating power amount prediction technique based on mixed model | |
| CN117390550A (en) | Low-carbon park carbon emission dynamic prediction method and system considering emission training set | |
| CN117134315A (en) | Distribution transformer load prediction method and device based on BERT algorithm | |
| CN116404637A (en) | Short-term load prediction method and device for electric power system | |
| CN113128754A (en) | GRU neural network-based residential water use prediction system and prediction method | |
| Jiahui et al. | Short-term load forecasting based on GA-PSO optimized extreme learning machine | |
| Cao et al. | Probabilistic electricity demand forecasting with transformer-guided state space model | |
| CN114444763A (en) | Wind power prediction method based on AFSA-GNN |
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 |