CN113988436B - Power consumption prediction method based on LSTM neural network and hierarchical relation correction - Google Patents
Power consumption prediction method based on LSTM neural network and hierarchical relation correction Download PDFInfo
- Publication number
- CN113988436B CN113988436B CN202111281571.3A CN202111281571A CN113988436B CN 113988436 B CN113988436 B CN 113988436B CN 202111281571 A CN202111281571 A CN 202111281571A CN 113988436 B CN113988436 B CN 113988436B
- Authority
- CN
- China
- Prior art keywords
- prediction
- power consumption
- electricity consumption
- user
- electricity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000012937 correction Methods 0.000 title claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 13
- 230000005611 electricity Effects 0.000 claims abstract description 78
- 230000003595 spectral effect Effects 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 38
- 230000015654 memory Effects 0.000 description 21
- 230000006399 behavior Effects 0.000 description 17
- 230000006872 improvement Effects 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000014509 gene expression Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- CIWBSHSKHKDKBQ-JLAZNSOCSA-N Ascorbic acid Chemical compound OC[C@H](O)[C@H]1OC(=O)C(O)=C1O CIWBSHSKHKDKBQ-JLAZNSOCSA-N 0.000 description 4
- 238000003064 k means clustering Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000002902 bimodal effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 239000002699 waste material Substances 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
- G06F18/2323—Non-hierarchical techniques based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Discrete Mathematics (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Operations Research (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (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 power consumption prediction method based on LSTM neural network and hierarchical relation correction, which belongs to the field of power consumption prediction and comprises the following steps: constructing a prediction variable system according to the electricity behavior characteristics of the user, the meteorological characteristics and the similar daily characteristics; clustering is carried out aiming at the electricity utilization behavior characteristics of the users by using a spectral clustering method, and a user classification result is obtained; short-term prediction is respectively carried out on the electricity consumption of different types of users based on the LSTM model; and correcting the prediction result according to the power consumption level relation to obtain the prediction result. The invention carries out short-term prediction of electricity consumption based on the LSTM model, and is used for carrying out short-term electricity consumption prediction on each type of clustered users; and correcting the prediction result according to the electricity consumption level relation, incorporating the regional electricity consumption data, correcting the individual electricity consumption prediction, and improving the prediction accuracy of the regional electricity consumption.
Description
Technical Field
The invention belongs to the field of electricity consumption prediction, and particularly relates to an electricity consumption prediction method based on LSTM neural network and hierarchical relation correction.
Background
The accurate electricity consumption prediction has positive effects on the normal operation of the power system and also has important effects on the analysis of macroscopic economy in the power industry. For the power producer, accurate power consumption prediction is convenient for the power producer to provide proper power supply, avoids the influence of insufficient power supply on life production, and also avoids power supply waste. Macroscopic economics and smart grid systems also require assessment of regional power industry development and economic analysis through power usage prediction. Nowadays, intelligent ammeter can real-time supervision user's electricity consumption action, and meteorological data also can be accurate to the region that the scope is very little in real time. In the big data age, the data capacity related to the electricity consumption is big, the data source is wide, and meanwhile, the electricity consumption prediction has great research significance, and the electricity consumption prediction becomes a subject of extensive research of domestic scholars.
The prediction and development of electricity consumption are mature method systems. The existing short-term prediction method of the electric quantity demand is mainly divided into a metering method, a fuzzy mathematic method and a machine learning method. The metering method comprises regression analysis, time sequence autocorrelation analysis and the like; the fuzzy mathematic method comprises a gray prediction method, a similar day method and the like; the machine learning method comprises a support vector machine, a random forest, a cyclic neural network, a long-term and short-term memory artificial neural network and the like. These prediction methods have been widely used in electricity consumption prediction, where machine learning methods can better identify nonlinear relationships between electricity consumption and other variables than metrology methods. Although the prediction method is mature, the existing research is less in consideration of the electricity behavior difference of different users, and model training is carried out by regarding all users as a whole, so that the prediction accuracy is reduced. Meanwhile, the existing research focuses on the prediction precision of the individual users, and the prediction of the regional power consumption only carries out simple addition on the prediction results of the individual users. This approach suffers from poor prediction accuracy at the macroscopic level.
Disclosure of Invention
The invention aims to provide a power consumption prediction method based on LSTM neural network and hierarchical relation correction.
In order to achieve the above purpose, the present invention provides a power consumption prediction method based on LSTM neural network and hierarchical relation correction, comprising the following steps:
1) Constructing a prediction variable system according to the electricity behavior characteristics of the user, the meteorological characteristics and the similar daily characteristics;
2) Clustering is carried out aiming at the electricity utilization behavior characteristics of the users by using a spectral clustering method, and a user classification result is obtained;
3) Short-term prediction is respectively carried out on the electricity consumption of different types of users based on the LSTM model;
4) And correcting the prediction result according to the power consumption level relation to obtain the prediction result.
As a further technical improvement, the constructed prediction variable body at least comprises user electricity utilization behavior characteristics, meteorological characteristics and similar day characteristics.
As a further technical improvement, the user electricity consumption behavior characteristics at least comprise historical electricity consumption, annual electricity consumption and monthly electricity consumption of the user, recent hour electricity consumption of the user, holiday electricity consumption of the user and early morning electricity consumption of the user.
As a further technical improvement, the meteorological features include air temperature, humidity, wind speed, wind direction, air pressure, and rainfall probability. When the model is constructed, the meteorological data is based on a history record; during model prediction, the weather data is based on the weather prediction result.
As a further technical improvement, the similar day characteristic comprises the same hours of electricity consumption before one day, two days, one week and one year.
The user electricity behavior characteristics are used for carrying out spectral clustering on the user, and the electricity behavior is roughly divided into: unimodal, bimodal, multimodal, stationary, four types; therefore, the present invention selects the cluster number of clusters to be 4. The clustering method selected by the invention is spectral clustering, which is a modern clustering algorithm, and has the advantage of high stability compared with the traditional K-means clustering method; the specific method of spectral clustering is as follows:
spectral clustering is a clustering algorithm derived according to the principle of graph theory, and is essentially to reduce the dimension of features through the feature vector of the graph Laplace matrix, so that different categories are more obviously classified. The specific algorithm is as follows:
assume that the sample is { x 1 ,…,x n X, where x i ∈R d ；
Step 1: constructing a similarity matrix S of the sample; defining a similarity matrix using a full join method, wherein
Step 2: construct adjacency matrix w=s of the graph and degree matrix d=diag (D) 1 ，…,d n ) Wherein
Step 3: the laplacian matrix of the graph is constructed.
The invention selects a symmetric standardized Laplace matrix, and the formula is as follows:
step 4: performing feature decomposition on the Laplace matrix:
L sym u i ＝λ i u i ,i＝1,…,n
wherein, lambda is more than or equal to 0 1 ≤…≤λ k ≤…≤λ n Constructing a matrix U= [ U ] 1 ,…,u k ]：
Step 5: normalizing the rows of the matrix U; building a matrix T E R n×k Elements of
Step 6: and taking the matrix T as a sample matrix, taking rows as samples, listing as characteristics, and performing K-means clustering to obtain a clustering result.
As a further technical improvement, the method for utilizing spectral clustering clusters aiming at the electricity utilization behavior characteristics of the users, and finally, the user group is divided into K categories, and each category of users has similar electricity utilization characteristics.
As a further technical improvement, the result of the spectral clustering is used for carrying out short-term prediction on the power consumption of different types of users based on the LSTM model to construct K types of LSTM prediction models, and each model is constructed as follows:
the LSTM prediction model takes the user line characteristics input at the current moment as information, weather characteristic information, similar day characteristic information and historic memory user behavior characteristic information, and the weather characteristic information and the similar day characteristic information output updated information through the functions of an input gate, a forget gate and an output gate; assuming that the input at time t is X t ∈R n×d The hidden state at the t-th moment is H t ∈R n×h The memory state at time t is C t ∈R n×h The method comprises the steps of carrying out a first treatment on the surface of the The input gate is I t Forgetful door F t And the output gate is O t The expressions of (2) are respectively:
I t ＝σ(X t W xi +H t-1 W hi +b i )
F t ＝σ(X t W xf +H t-1 W hf +b f )
O t ＝σ(X t W xo +H t-1 W ho +b o )
wherein W is xi ,W xf ,W xo ,W hi ,W hf ,W ho Is a weight parameter, b i ,b f ,b o Is a bias parameter, σ () represents a sigmoid function;
updating the hidden state at the t moment needs to be updated through the memory unit; the current state input characteristics and the information of the previous state hiding layer pass through a memory gate to obtain a potential memory unit of the current state:
next, the information of the input gate, the forget gate and the output gate is used for memorizing the information, hiding the state and pre-processingUpdating the measurement result; memory cell C at time t t From the last state memory cell C t-1 Potential memory unit for restoring current state through forgetting gate
The expression is as follows:
wherein W is xc Is the weight parameter of the current input acting on the potential memory unit, W hc Is the weight parameter of the hidden state to the potential memory unit at the last moment, b c Representing bias, as would be the corresponding multiplication of elements between the matrices; updating the hidden layer in the current state is obtained through an output gate, and the expression is as follows:
H t ＝O t ⊙tanh(C t )
the prediction result of the current state is updated as follows:
y t ＝H t W hy +b y
wherein W is hy Weight vector representing prediction result of hidden layer, b y Representing the bias term.
As a further technical improvement, the step of correcting the prediction result according to the power consumption level relation is as follows:
step 1: obtaining a prediction result by using an LSTM model
Wherein the method comprises the steps of
Predictive result representing the amount of electricity used by the user, +.>
Prediction of the power consumption of the representation area;
step 2: correcting a prediction result; recording corrected individual user prediction results
Find the optimal correction matrix +.>
So that the corrected individual user prediction result +.>
Meanwhile, the corrected regional power consumption prediction is equal to the sum of the power consumption predictions of the individual users under the jurisdiction; minimizing the modified prediction result +.>
Is a variance of (2);
step 3: obtaining corrected overall prediction results:
as a further technical improvement, the method finds the optimal correction matrix
Taking the variance of the minimum corrected overall prediction error as a criterion; error before correction>
The variance is->
By minimizing the variance of the corrected overall error:
obtaining an optimal weight matrix:
as a further technical improvement, the variance matrix W h In the actual electricity consumption prediction, estimation is needed, and the estimation method is as follows: combining the characteristic that the short-term electricity consumption prediction usually takes half an hour as a time interval, taking
Where s= (t+h) mod48, I () represents an indicator function.
Compared with the prior art, the invention has the following beneficial technical effects:
the invention carries out short-term prediction of electricity consumption based on the LSTM model, and is used for carrying out short-term electricity consumption prediction on each type of clustered users; and correcting the prediction result according to the electricity consumption level relation, incorporating the regional electricity consumption data, correcting the individual electricity consumption prediction, and improving the prediction accuracy of the regional electricity consumption.
Drawings
In order to more clearly illustrate the embodiments of the present invention, the drawings required for the embodiments will be briefly described below, and it will be apparent that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be derived from the drawings before the inventive work is not performed for a person of ordinary skill in the art.
Fig. 1 is a schematic flow chart of a power consumption prediction model based on a hierarchical aggregation relationship.
Detailed Description
Embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
Other advantages and effects of the present disclosure will become readily apparent to those skilled in the art from the following disclosure, which describes embodiments of the present disclosure by way of specific examples. It will be apparent that the described embodiments are merely some, but not all embodiments of the present disclosure. The disclosure may be embodied or practiced in other different specific embodiments, and details within the subject specification may be modified or changed from various points of view and applications without departing from the spirit of the disclosure. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.
Example 1
As shown in fig. 1, a power consumption prediction method based on LSTM neural network and hierarchical relation correction is characterized by comprising the following steps:
1) Constructing a prediction variable system according to the electricity behavior characteristics of the user, the meteorological characteristics and the similar daily characteristics;
2) Clustering is carried out aiming at the electricity utilization behavior characteristics of the users by using a spectral clustering method, and a user classification result is obtained;
3) Short-term prediction is respectively carried out on the electricity consumption of different types of users based on the LSTM model;
4) And correcting the prediction result according to the power consumption level relation to obtain the prediction result.
Example two
One difference from the embodiment is that: the construction prediction variable body at least comprises user electricity behavior characteristics, meteorological characteristics and similar day characteristics.
The user electricity consumption behavior characteristics at least comprise historical electricity consumption, annual electricity consumption and monthly electricity consumption of the user, recent hour electricity consumption of the user, holiday electricity consumption of the user and early morning electricity consumption of the user.
The meteorological features include air temperature, humidity, wind speed, wind direction, air pressure, and rainfall probability.
The similar day characteristic includes the same hours of electricity usage one day ago, two days ago, one week ago, and one year ago.
Example III
The second difference from the embodiment is that:
the user electricity behavior characteristics are used for carrying out spectral clustering on the user, and the electricity behavior is roughly divided into: unimodal, bimodal, multimodal, stationary, four types; therefore, the present invention selects the cluster number of clusters to be 4. The clustering method selected by the invention is spectral clustering, which is a modern clustering algorithm, and has the advantage of high stability compared with the traditional K-means clustering method; the specific method of spectral clustering is as follows:
spectral clustering is a clustering algorithm derived according to the principle of graph theory, and is essentially to reduce the dimension of features through the feature vector of the graph Laplace matrix, so that different categories are more obviously classified. The specific algorithm is as follows:
assume that the sample is { x 1 ,…,x n X, where x i ∈R d ；
Step 1: constructing a similarity matrix S of the sample; defining a similarity matrix using a full join method, wherein
Step 2: construct adjacency matrix w=s of the graph and degree matrix d=diag (D) 1 ，…,d n ) Wherein
Step 3: the laplacian matrix of the graph is constructed.
The invention selects a symmetric standardized Laplace matrix, and the formula is as follows:
step 4: performing feature decomposition on the Laplace matrix:
L sym u i ＝λ i u i ,i＝1,…,n
wherein, lambda is more than or equal to 0 1 ≤…≤λ k ≤…≤λ n Constructing a matrix U= [ U ] 1 ,…,u k ]：
Step 5: normalizing the rows of the matrix U; constructionMatrix T epsilon R n×k Elements of
Step 6: and taking the matrix T as a sample matrix, taking rows as samples, listing as characteristics, and performing K-means clustering to obtain a clustering result.
The method for utilizing spectral clustering aims at the electricity utilization behavior characteristics of users to cluster, and finally, the user group is divided into K categories, and each category of users has similar electricity utilization characteristics.
And carrying out power consumption prediction, wherein short-term prediction of different types of user power consumption is carried out based on an LSTM model, a K type LSTM prediction model is constructed by using the result of spectral clustering, and each model is constructed as follows: the LSTM prediction model takes the user line characteristics input at the current moment as information, weather characteristic information, similar day characteristic information and historic memory user behavior characteristic information, and the weather characteristic information and the similar day characteristic information output updated information through the functions of an input gate, a forget gate and an output gate; the LSTM model has excellent performance in long-time sequence prediction, and the LSTM selectively memorizes or ignores information of hidden states at a previous time point by a gating method. The mechanism is determined by an input gate, a forget gate and an output gate, and the input at the t-th moment is assumed to be X t ∈R n×d The hidden state at the t-th moment is H t ∈R n×h The memory state at time t is C t ∈R n×h The method comprises the steps of carrying out a first treatment on the surface of the The input gate is I t Forgetful door F t And the output gate is O t The expressions of (2) are respectively:
I t ＝σ(X t W xi +H t-1 W hi +b i )
F t ＝σ(X t W xf +H t-1 W hf +b f )
O t ＝σ(X t W xo +H t-1 W ho +b o )
wherein W is xi ,W xf ,W xo ,W hi ,W hf ,W ho Is a weight parameter, b i ,b f ,b o Is a bias parameter, σ () represents a sigmoid function;
updating the hidden state at the t moment needs to be updated through the memory unit; the current state input characteristics and the information of the previous state hiding layer pass through a memory gate to obtain a potential memory unit of the current state:
next, updating the memory information, the hidden state and the prediction result by using the information of the input gate, the forget gate and the output gate; memory cell C at time t t From the last state memory cell C t-1 Potential memory unit for restoring current state through forgetting gate
The expression is as follows:
wherein W is xc Is the weight parameter of the current input acting on the potential memory unit, W hc Is the weight parameter of the hidden state to the potential memory unit at the last moment, b c Representing bias, as would be the corresponding multiplication of elements between the matrices; updating the hidden layer in the current state is obtained through an output gate, and the expression is as follows:
H t ＝O t ⊙tanh(C t )
the prediction result of the current state is updated as follows:
y t ＝H t W hy +b y
wherein W is hy Weight vector representing prediction result of hidden layer, b y Representing the bias term.
And carrying out prediction result correction according to the power utilization level relation, wherein the prediction result correction comprises two processes of level relation and prediction result correction. The hierarchical relationship is as follows:suppose C j Representing an area (a village, a district, a city or an entire autonomous area),
indicating the amount of electricity used in the area. The following relationship holds: />
Integrating all individual electricity consumption into one vector Y t ＝[y 1t ,…,y Nt ] T . Integrating all individual power consumption and regional power consumption into a vector
Then, the following relationship holds: z is Z t ＝SY t . Where S is a matrix composed of 0 and 1 as elements.
The prediction result correction is performed based on a hierarchical relationship, and the correction of the prediction result cannot change the relationship that the regional power consumption is equal to the sum of the power consumption of the users managed by the region. One simple idea is to sum individual predictions through jurisdictions, but this introduces large errors into regional power usage predictions. The invention adopts the method that the LSTM model is respectively and independently established for the individual user and the regional power consumption to obtain the prediction result
(future h-phase). At this time, a->
No longer satisfies, and needs to be corrected, the correction process is as follows:
prediction of power usage for individual users
Through the whole prediction result->
I.e. V. Modified global prediction result->
The method meets the following conditions: />
This process satisfies the relationships between levels without losing much prediction accuracy. The key process of correction is to construct a weight matrix P h The construction process is as follows:
constructing a weight matrix P by taking the minimized variance of the corrected overall prediction error as a criterion h . Total error before correction
The variance is->
By minimizing the variance of the corrected overall error: />
Obtaining an optimal weight matrix:
obtaining corrected overall prediction results:
the error variance W before correction is needed in the actual power consumption prediction h The estimation is carried out, and the estimation method is as follows:
combining the characteristic that the short-term electricity consumption prediction usually takes half an hour as a time interval, taking
Where s= (t+h) mod48, I () represents an indicator function.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
Claims (4)
1. The power consumption prediction method based on LSTM neural network and hierarchical relation correction is characterized by comprising the following steps of:
1) Constructing a prediction variable system according to the electricity behavior characteristics, the meteorological characteristics and the similar daily characteristics of the user;
2) Clustering is carried out aiming at the electricity utilization behavior characteristics of the users by using a spectral clustering method, the user group is divided into K categories, and each category of users has similar electricity utilization behavior characteristics to obtain user classification results;
3) Short-term prediction is respectively carried out on the electricity consumption of different types of users based on the LSTM prediction model;
4) Carrying out prediction result correction according to the power consumption level relation to obtain a corrected prediction result;
the method for carrying out short-term prediction on the power consumption of different types of users based on the LSTM prediction model comprises the following steps: constructing a K-type LSTM prediction model by using the result of the spectral clustering;
the step of correcting the prediction result according to the power consumption level relation is as follows:
step 1: the LSTM prediction model obtains the whole prediction result
Wherein->
Representing the predicted result of the user's power consumption,/-, and>
indicating regional power usage pre-allocationMeasuring results;
step 2: correcting a prediction result; recording the corrected user electricity consumption prediction result
Searching for an optimal correction matrix P * So that the corrected prediction result of the user's power consumption +.>
Meanwhile, the corrected regional power consumption prediction result is equal to the sum of the power consumption prediction results of the individual users under the jurisdiction;
step 3: obtaining corrected overall prediction results:
wherein, searching the optimal correction matrix P * Taking the variance of the error of the minimum corrected overall prediction result as a criterion; recording errors of the overall prediction results before correction
The variance matrix is->
By minimizing the variance of the error of the corrected overall prediction result:
obtaining an optimal correction matrix: p (P) * ＝(S T W -1 S) -1 S T W -1 ；
The variance matrix W is obtained through estimation in the actual electricity consumption prediction, and the estimation method is as follows: taking at a time interval of half an hour
Where s= (t+h) mod48, I () represents an indicator function.
2. The power consumption prediction method based on LSTM neural network and hierarchical relation correction according to claim 1, wherein: the user electricity consumption behavior characteristics comprise historical electricity consumption of a user, annual electricity consumption, monthly electricity consumption, recent hour electricity consumption of the user, holiday electricity consumption of the user and early morning electricity consumption of the user.
3. The power consumption prediction method based on LSTM neural network and hierarchical relation correction according to claim 1, wherein: the meteorological features include air temperature, humidity, wind speed, wind direction, air pressure and rainfall likelihood.
4. The power consumption prediction method based on LSTM neural network and hierarchical relation correction according to claim 2, wherein: the similar day characteristic includes the same hours of electricity usage one day ago, two days ago, one week ago, and one year ago.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111281571.3A CN113988436B (en) | 2021-11-01 | 2021-11-01 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111281571.3A CN113988436B (en) | 2021-11-01 | 2021-11-01 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113988436A CN113988436A (en) | 2022-01-28 |
| CN113988436B true CN113988436B (en) | 2023-04-28 |
Family
ID=79745265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111281571.3A Active CN113988436B (en) | 2021-11-01 | 2021-11-01 | Power consumption prediction method based on LSTM neural network and hierarchical relation correction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113988436B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115660225A (en) * | 2022-12-13 | 2023-01-31 | 浙江万胜智能科技股份有限公司 | Electricity load prediction management method and system based on ammeter communication module |
| CN116386215B (en) * | 2023-03-16 | 2024-04-19 | 淮阴工学院 | Intelligent charging method for mobile electric box based on people flow |
| CN117787652A (en) * | 2024-01-11 | 2024-03-29 | 湖北华中电力科技开发有限责任公司 | Park power demand prediction method, related device and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109063911A (en) * | 2018-08-03 | 2018-12-21 | 天津相和电气科技有限公司 | A kind of Load aggregation body regrouping prediction method based on gating cycle unit networks |
| CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
| CN110674993A (en) * | 2019-09-26 | 2020-01-10 | 广东电网有限责任公司 | User load short-term prediction method and device |
| CN112330028A (en) * | 2020-11-08 | 2021-02-05 | 国网天津市电力公司 | Electric bus charging load prediction method based on spectral clustering and LSTM neural network |
| CN112561156A (en) * | 2020-12-11 | 2021-03-26 | 国网江苏省电力有限公司南通供电分公司 | Short-term power load prediction method based on user load mode classification |
| CN113033596A (en) * | 2020-12-30 | 2021-06-25 | 国网河南省电力公司南阳供电公司 | Refined identification method for user electricity consumption behavior category and typical electricity consumption mode |
| CN113255900A (en) * | 2021-06-23 | 2021-08-13 | 河北工业大学 | Impulse load prediction method considering improved spectral clustering and Bi-LSTM neural network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11025061B2 (en) * | 2019-03-28 | 2021-06-01 | Accenture Global Solutions Limited | Predictive power usage monitoring |
| CN110298501B (en) * | 2019-06-21 | 2022-08-16 | 河海大学常州校区 | Electrical load prediction method based on long-time and short-time memory neural network |
| CN112651543A (en) * | 2020-11-10 | 2021-04-13 | 沈阳工程学院 | Daily electric quantity prediction method based on VMD decomposition and LSTM network |
-
2021
- 2021-11-01 CN CN202111281571.3A patent/CN113988436B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109063911A (en) * | 2018-08-03 | 2018-12-21 | 天津相和电气科技有限公司 | A kind of Load aggregation body regrouping prediction method based on gating cycle unit networks |
| CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
| CN110674993A (en) * | 2019-09-26 | 2020-01-10 | 广东电网有限责任公司 | User load short-term prediction method and device |
| CN112330028A (en) * | 2020-11-08 | 2021-02-05 | 国网天津市电力公司 | Electric bus charging load prediction method based on spectral clustering and LSTM neural network |
| CN112561156A (en) * | 2020-12-11 | 2021-03-26 | 国网江苏省电力有限公司南通供电分公司 | Short-term power load prediction method based on user load mode classification |
| CN113033596A (en) * | 2020-12-30 | 2021-06-25 | 国网河南省电力公司南阳供电公司 | Refined identification method for user electricity consumption behavior category and typical electricity consumption mode |
| CN113255900A (en) * | 2021-06-23 | 2021-08-13 | 河北工业大学 | Impulse load prediction method considering improved spectral clustering and Bi-LSTM neural network |
Non-Patent Citations (1)
| Title |
|---|
| 徐冰涵 等.考虑分时电价的居民用户短期用电量分类预测及修正方法.《电力***保护与控制》.2020,第48卷(第06期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113988436A (en) | 2022-01-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113988436B (en) | Power consumption prediction method based on LSTM neural network and hierarchical relation correction | |
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN112508275B (en) | Power distribution network line load prediction method and equipment based on clustering and trend indexes | |
| CN108280551B (en) | Photovoltaic power generation power prediction method utilizing long-term and short-term memory network | |
| CN111260136A (en) | Building short-term load prediction method based on ARIMA-LSTM combined model | |
| CN111027772A (en) | Multi-factor short-term load prediction method based on PCA-DBILSTM | |
| CN110503256A (en) | Short-term load forecasting method and system based on big data technology | |
| CN113705877A (en) | Real-time monthly runoff forecasting method based on deep learning model | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN113537469B (en) | Urban water demand prediction method based on LSTM network and Attention mechanism | |
| CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
| CN114862032B (en) | XGBoost-LSTM-based power grid load prediction method and device | |
| CN115115125A (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
| CN115796004A (en) | Photovoltaic power station ultra-short term power intelligent prediction method based on SLSTM and MLSTNet models | |
| CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
| CN113011680A (en) | Power load prediction method and system | |
| CN114595861A (en) | MSTL (modeling, transformation, simulation and maintenance) and LSTM (least Square TM) model-based medium-and-long-term power load prediction method | |
| CN116523142A (en) | Photovoltaic cluster interval prediction method considering space-time characteristic clustering | |
| CN116757057A (en) | Air quality prediction method based on PSO-GA-LSTM model | |
| CN113537582A (en) | Photovoltaic power ultra-short-term prediction method based on short-wave radiation correction | |
| CN115759415A (en) | Power consumption demand prediction method based on LSTM-SVR | |
| Li et al. | Performance degradation decomposition-ensemble prediction of PEMFC using CEEMDAN and dual data-driven model | |
| CN114897264A (en) | Photovoltaic output interval prediction method under small sample scene based on transfer learning | |
| CN117034046A (en) | Flexible load adjustable potential evaluation method based on ISODATA clustering | |
| CN116011655A (en) | Load ultra-short-term prediction method and system based on two-stage intelligent feature engineering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |