CN113705915A - CNN-LSTM-ARIMA-based combined short-term power load prediction method - Google Patents
CNN-LSTM-ARIMA-based combined short-term power load prediction method Download PDFInfo
- Publication number
- CN113705915A CN113705915A CN202111018225.6A CN202111018225A CN113705915A CN 113705915 A CN113705915 A CN 113705915A CN 202111018225 A CN202111018225 A CN 202111018225A CN 113705915 A CN113705915 A CN 113705915A
- Authority
- CN
- China
- Prior art keywords
- data
- cnn
- lstm
- arima
- model
- 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 25
- 238000012549 training Methods 0.000 claims abstract description 22
- 230000007774 longterm Effects 0.000 claims abstract description 5
- 230000004913 activation Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 12
- 238000002790 cross-validation Methods 0.000 claims description 10
- 238000013277 forecasting method Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008034 disappearance Effects 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 abstract description 3
- 239000000284 extract Substances 0.000 abstract description 2
- 238000012795 verification Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000000306 recurrent effect 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
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- 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)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of short-term load prediction of a power system, and discloses a CNN-LSTM-ARIMA-based combined short-term power load prediction method. The CNN in the combined model is responsible for extracting the short-term mode in the time dimension and the local dependency relationship between variables, the LSTM extracts the long-term correlation information of the load data, and the ARIMA is used as a linear component to be fused with the prediction result of the main model. The method comprises the following steps: selecting and determining a basic structure and related network parameters of the combined network model; obtaining sample data and importing the sample data into a combined model for training; and selecting an optimal prediction model by using a K-fold cross verification method. The invention aims to reduce the daily load prediction error rate by using small and medium-scale load data, thereby providing safe and reliable data support for power dispatching work. Compared with the prior art, the method has the advantages of improving the short-term power load prediction precision and reducing the training time.
Description
Technical Field
The invention belongs to the technical field of power systems, and particularly relates to a CNN-LSTM-ARIMA combination-based short-term power load prediction model building method.
Background
The invention belongs to the technical field of power systems, and particularly relates to a CNN-LSTM-ARIMA combination-based short-term power load prediction model building method.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a CNN-LSTM-ARIMA-based combined short-term power load prediction method, which is used for establishing a novel prediction model by combining K-fold cross validation and a deep neural network, can realize accurate prediction of short-term power load data by using medium and small-scale historical load data sets, and aims to reduce the short-term load prediction error rate so as to provide data support for the safe scheduling of a power grid system.
The purpose of the invention is realized as follows: a combined short-term power load forecasting method based on CNN-LSTM-ARIMA is applied to power load data to complete a short-term power load forecasting task and comprises the following steps:
step 1: selecting and determining a basic structure and related network parameters of a CNN-LSTM-ARIMA combined model;
step 2: acquiring power load data, and performing data preprocessing including vacancy value filling and data normalization;
wherein, the vacancy value is filled: filling the vacancy value by using the data mean value of two adjacent time points; to ensure the integrity of the power load data on the timeline,
data normalization: mapping data to [0,1 ]]In the interval, the normalization formula is:in the formula:representing normalized data values, x (i) representing raw data of variables, xmaxRepresents the maximum in the raw data; x is the number ofminRepresents the minimum in the raw data;
and step 3: dividing a data set by adopting a k-fold cross validation method, and taking Mean Square Error (MSE) as an error function;
and 4, step 4: constructing input by using a sliding window method, setting the window width to be 24, and importing a data set into a CNN-LSTM-ARIMA combined model;
and 5: recording the generalization error obtained by each training, wherein the formula is as follows:
step 6: adjusting the parameter setting of the network model, training for N times, and evaluating to obtain the error index of each modelAnd comparing the MSE to find a super-parameter value which enables the generalization performance of the model to be optimal, and selecting the optimal model for deployment.
Further, the first layer of the CNN-LSTM-ARIMA combined prediction model is a convolutional neural network CNN with pooling removed, the convolution parameters are set to filters 48, kernel _ size 6, strings 1, activation 'return', the convolutional layer is composed of a plurality of filters with width ω and height n, and the k-th filter sweeps the input matrix X and generates hkThe formula is as follows: h isk=RELU(ωk*X+bk);
In the formula: is convolution operation and outputs hkFor implicit vectors, RELU activation function
Furthermore, the output of the CNN-LSTM-ARIMA combined prediction model convolution layer is simultaneously input into a recursive component and a recursive jump component, and the recursive component and the recursive jump component adopt an LSTM network structure and can capture long-term correlation information among power load data;
the recursion component is a recursion layer of the gated recursion unit, a ReLU activation function is used as an implicit updating activation function, and the hidden state calculation formula of the recursion unit at the time t is as follows:
rt=σ(xtWxr+ht-1Whr+br)
ut=σ(xtWxu+ht-1Whu+bu)
in the formula:is the product of elements, sigma is the sigmoid function, xtThe input of the layer at the time t, and the output of the layer is the hidden state of each time step;
the recursive skip component is to add skip-links between the current hidden unit and the hidden unit in the same phase in the adjacent cycle, which can alleviate the problem of gradient disappearance, and the update process can be expressed as follows:
rt=σ(xtWxr+ht-pWhr+br)
ut=σ(xtWxu+ht-pWhu+bu)
where the input to the layer is the output of the convolution component and p is the number of hidden cells skipped.
Further, the CNN-LSTM-ARIMA combined prediction model combines the outputs of the recursive component and the recursive skip component components by a fully connected layer, the inputs of the fully connected layer including the hidden state of the recurrentcomponent at timestamp t, the symbolic representations of p hidden states from timestamp t-p +1 to the recurrentskip component at timestamp t, the output of the fully connected layer calculated as:
wherein the content of the first and second substances,is the predicted result of the model at the time t.
Further, an autoregressive model is adopted as a linear component, and the autoregressive branch model represents that the prediction result of the autoregressive component is as follows:
and superposing the main model output and the autoregressive component output and combining a sigmoid function to obtain a final power load prediction as follows:
furthermore, the short-term load prediction method provided by the invention can ensure higher prediction accuracy on a medium-small-scale data set, the data set is divided into K parts in an equal proportion through K-fold cross validation, one part of the K parts is used as test data, the other K-1 parts of the K-fold cross validation are used as training data, and each sample data has the chance of being divided into the training set or the test set only once in each iteration process.
Further, the training sample data comprises an input sample and an output sample, the output sample is a historical power load value, and the input sample comprises temperature information, type information, climate information and historical daily load information of a sample day.
Further, in step 2, before the power load data sample is used as a training sample, preprocessing operations of outlier elimination, vacancy filling and data normalization are performed.
Compared with the prior art, the invention has the outstanding and beneficial technical effects that: the CNN-LSTM-ARIMA combined prediction model provided by the invention can ensure higher prediction accuracy on medium and small-scale historical load data sets. The data set is divided into K parts in equal proportion, each experiment selects one different data part from the K parts as test data (the data of the K parts are ensured to be respectively subjected to test data), the rest K-1 data are taken as training data, and finally the obtained K experiment results are divided equally. The K-fold cross validation can acquire effective information as much as possible from limited data and can effectively reduce overfitting.
The CNN-LSTM-ARIMA combined prediction model provided by the invention utilizes the convolution layer to effectively extract the short-term mode of the load data in the time dimension and the local dependency relationship between variables, the recursive component and the recursive jump component capture the long-term correlation information between the power load data, and the ARIMA linear component is added. Therefore, the associated information and the variation trend in the power load data set are comprehensively mined, and a good prediction model is established.
Accurate short-term load prediction can provide data support for power grid dispatching, power grid power supply reliability is improved, the risk of unplanned power failure is reduced, and national economic level improvement is promoted. Accurate short-term load prediction can guarantee the balance of supply and demand of an energy network and promote the construction of a smart power grid.
Drawings
FIG. 1 is a diagram of the CNN-LSTM-ARIMA model prediction step;
FIG. 2 is a schematic diagram of the CNN-LSTM-ARIMA algorithm;
FIG. 3 is a CNN-LSTM-ARIMA network architecture nonlinear element;
FIG. 4 is a comparison of a predicted curve and an actual curve for a weekday;
FIG. 5 is a comparison of predicted versus actual curves for Wednesday;
FIG. 6 shows MAPE and RMSE for each algorithm of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the following description, with reference to the drawings in the embodiments of the present invention, clearly and completely describes the technical solution in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-6, the invention provides a combined short-term power load forecasting method based on CNN-LSTM-ARIMA, which is applied to power load data to complete a short-term power load forecasting task, and comprises the following steps:
step 1: selecting and determining a basic structure and related network parameters of a CNN-LSTM-ARIMA combined model;
step 2: acquiring power load data, and performing data preprocessing including vacancy value filling and data normalization;
wherein, the vacancy value is filled: filling the vacancy value by using the data mean value of two adjacent time points; to ensure the integrity of the power load data on the timeline,
data normalization: mapping data to [0,1 ]]In the interval, the normalization formula is:in the formula:representing normalized data values, x (i) representing raw data of variables, xmaxRepresents the maximum in the raw data; x is the number ofminRepresents the minimum in the raw data;
and step 3: dividing a data set by adopting a k-fold cross validation method, and taking Mean Square Error (MSE) as an error function;
and 4, step 4: constructing input by using a sliding window method, setting the window width to be 24, and importing a data set into a CNN-LSTM-ARIMA combined model;
schematic diagram of CNN-LSTM-ARIMA algorithm is shown in FIG. 2Shown in the figure. The first layer of the CNN-LSTM-ARIMA combined prediction model is a convolutional neural network CNN with a pooling layer removed, and the short-term mode of load data in a time dimension and the local dependency relationship among variables can be effectively extracted. The convolution parameters are set to filters 48, kernel _ size 6, threads 1, activation relu', the convolution layer is composed of a plurality of filters with width omega and height n (the height is set to be consistent with the number of variables), the k-th filter sweeps the input matrix X and generates hkThe formula is as follows: h isk=RELU(ωk*X+bk);
In the formula: is convolution operation and outputs hkFor implicit vectors, RELU activation function
The output of the CNN-LSTM-ARIMA combined prediction model convolution layer is simultaneously input into a recursion Component (recursion Component) and a recursion Skip Component (recursion Skip Component), and the recursion Component and the recursion Skip Component adopt an LSTM network structure and can capture long-term related information among power load data; such as the periodicity of the load data. If the experimental operation environment supports GPU acceleration, CuDNNLSTM is used when the model is proposed to be built, the training speed can be obviously increased by about ten times compared with the traditional LSTM.
The recursion component is a recursion layer of the gated recursion unit, a ReLU activation function is used as an implicit updating activation function, and the hidden state calculation formula of the recursion unit at the time t is as follows:
rt=σ(xtWxr+ht-1Whr+br)
ut=σ(xtWxu+ht-1Whu+bu)
in the formula:is the product of elements, sigma is the sigmoid function, xtThe input of the layer at the time t, and the output of the layer is the hidden state of each time step;
the recursive skip component is to add skip-links between the current hidden unit and the hidden unit in the same phase in the adjacent cycle, which can alleviate the problem of gradient disappearance, and the update process can be expressed as follows:
rt=σ(xtWxr+ht-pWhr+br)
ut=σ(xtWxu+ht-pWhu+bu)
wherein the input to the layer is the output of the convolution component, p is the number of skipped hidden cells;
the CNN-LSTM-ARIMA combined prediction model combines the outputs of the recursive component and the recursive skip component components by a fully connected layer (Dense), the inputs of which include the hidden state of the Recurrent component at timestamp t (htR), the symbolic representations of the p hidden states from timestamp t-p +1 to the Recurrent-skip component at timestamp t, the output of the fully connected layer is calculated as:
wherein the content of the first and second substances,the predicted result of the model at the time t is obtained;
one major drawback of neural network models is that the scale of the output is not sensitive to the scale of the input due to the non-linear nature of the convolution and recursion components. In order to solve the defect that the scale of an input signal is changed non-periodically continuously, the prediction precision of a neural network model is obviously reduced, a classical autoregressive (ARIMA) model is used as a linear component, and all input dimensions are reserved by intercepting the time dimensions of nearly 3 windows. The autoregressive branch model represents the prediction result of the autoregressive component as follows:
and superposing the main model output and the autoregressive component output and combining a sigmoid function to obtain a final power load prediction as follows:
and 5: recording the generalization error obtained by each training, wherein the formula is as follows:
step 6: adjusting the parameter setting of the network model, training for N times, and evaluating to obtain the error index of each modelAnd comparing the MSE to find a super-parameter value which enables the generalization performance of the model to be optimal, and selecting the optimal model for deployment.
Furthermore, the short-term load prediction method provided by the invention can ensure higher prediction accuracy on a medium-small-scale data set, the data set is divided into K parts in an equal proportion through K-fold cross validation, one part of the K parts is used as test data, the other K-1 parts of the K-fold cross validation are used as training data, and each sample data has the chance of being divided into the training set or the test set only once in each iteration process.
Further, the training sample data comprises an input sample and an output sample, the output sample is a historical power load value, and the input sample comprises temperature information, type information, climate information, historical daily load and other information of a sample day.
Further, in step 2, before the power load data sample is used as a training sample, preprocessing operations of outlier elimination, vacancy filling and data normalization are performed.
The above embodiments are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereby, so: all equivalent changes made according to the structure, shape and principle of the invention are covered by the protection scope of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
Claims (8)
1. A CNN-LSTM-ARIMA combination short-term power load forecasting method is characterized in that the method is applied to power load data to complete a short-term power load forecasting task and comprises the following steps:
step 1: selecting and determining a basic structure and related network parameters of a CNN-LSTM-ARIMA combined model;
step 2: acquiring power load data, and performing data preprocessing including vacancy value filling and data normalization;
wherein, the vacancy value is filled: filling the vacancy value by using the data mean value of two adjacent time points; to ensure the integrity of the power load data on the timeline,
data normalization: mapping data to [0,1 ]]In the interval, the normalization formula is:in the formula:representing normalized data values, x (i) representing raw data of variables, xmaxRepresents the maximum in the raw data; x is the number ofminRepresents the minimum in the raw data;
and step 3: dividing a data set by adopting a k-fold cross validation method, and taking Mean Square Error (MSE) as an error function;
and 4, step 4: constructing input by using a sliding window method, setting the window width to be 24, and importing a data set into a CNN-LSTM-ARIMA combined model;
and 5: recording the generalization error obtained by each training, wherein the formula is as follows:
step 6: adjusting the parameter setting of the network model, training for N times, and evaluating to obtain the error index of each modelAnd comparing the MSE to find a super-parameter value which enables the generalization performance of the model to be optimal, and selecting the optimal model for deployment.
2. The CNN-LSTM-ARIMA combined short-term power load prediction method according to claim 1, wherein the first layer of the CNN-LSTM-ARIMA combined prediction model is a convolutional neural network CNN with a removed pooling layer, the convolution parameters are set to filters 48, kernel _ size 6, threads 1, and activation 'relu', the convolutional layer is composed of a plurality of filters with width ω and height n, and the k-th filter sweeps the input matrix X and generates hkThe formula is as follows: h isk=RELU(ωk*X+bk);
3. The CNN-LSTM-ARIMA combined short-term power load forecasting method as claimed in claim 1, wherein the output of the CNN-LSTM-ARIMA combined forecasting model convolution layer is simultaneously input to a recursive component and a recursive jump component, and the recursive component and the recursive jump component adopt an LSTM network structure and can capture long-term correlation information among power load data;
the recursion component is a recursion layer of the gated recursion unit, a ReLU activation function is used as an implicit updating activation function, and the hidden state calculation formula of the recursion unit at the time t is as follows:
rt=σ(xtWxr+ht-1Whr+br)
ut=σ(xtWxu+ht-1Whu+bu)
in the formula:is the product of elements, sigma is the sigmoid function, xtThe input of the layer at the time t, and the output of the layer is the hidden state of each time step;
the recursive skip component is to add skip-links between the current hidden unit and the hidden unit in the same phase in the adjacent cycle, which can alleviate the problem of gradient disappearance, and the update process can be expressed as follows:
rt=σ(xtWxr+ht-pWhr+br)
ut=σ(xtWxu+ht-pWhu+bu)
where the input to the layer is the output of the convolution component and p is the number of hidden cells skipped.
4. The CNN-LSTM-ARIMA based combined short-term power load forecasting method of claim 1, characterized in that the CNN-LSTM-ARIMA combined forecasting model combines the outputs of the recursive component and the recursive skip component with a full connection layer, the input of the full connection layer comprises the hidden state of the recurrentcomponent at timestamp t, the symbolic representations of p hidden states from timestamp t-p +1 to the recurrentskip component at timestamp t, the output of the full connection layer is calculated as:
5. The combined short-term electrical load forecasting method based on CNN-LSTM-ARIMA as claimed in claim 1, wherein an autoregressive model is adopted as a linear component, and the autoregressive branch model represents that the forecasting result of the autoregressive component is:
6. the CNN-LSTM-ARIMA combination short-term power load forecasting method as claimed in claim 1, wherein the short-term load forecasting method provided by the invention can ensure higher forecasting accuracy on medium and small-scale data sets, the data sets are divided into K parts in equal proportion through K-fold cross validation, one part of the K parts is used as test data, the other K-1 parts are used as training data, and each sample data has only one chance of being divided into the training set or the test set in each iteration process.
7. The CNN-LSTM-ARIMA based combined short term power load forecasting method of claim 1, wherein the training sample data includes an input sample and an output sample, the output sample is a historical power load value, and the input sample includes temperature information, type information, climate information and historical daily load information of a sample day.
8. The CNN-LSTM-ARIMA-based combined short-term power load forecasting method as claimed in claim 1, wherein in step 2, the power load data samples are pre-processed by outlier elimination, null filling and data normalization before being used as training samples.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111018225.6A CN113705915A (en) | 2021-09-01 | 2021-09-01 | CNN-LSTM-ARIMA-based combined short-term power load prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111018225.6A CN113705915A (en) | 2021-09-01 | 2021-09-01 | CNN-LSTM-ARIMA-based combined short-term power load prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113705915A true CN113705915A (en) | 2021-11-26 |
Family
ID=78658533
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111018225.6A Pending CN113705915A (en) | 2021-09-01 | 2021-09-01 | CNN-LSTM-ARIMA-based combined short-term power load prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113705915A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115146842A (en) * | 2022-06-24 | 2022-10-04 | 沈阳建筑大学 | Multivariate time series trend prediction method and system based on deep learning |
CN115329659A (en) * | 2022-07-18 | 2022-11-11 | 浙江大学 | Method and system for real-time early warning and intelligent control of dioxin emission in waste incinerator |
CN115860270A (en) * | 2023-02-21 | 2023-03-28 | 保定博堃元信息科技有限公司 | Network supply load prediction system and method based on LSTM neural network |
CN116227543A (en) * | 2023-05-09 | 2023-06-06 | 南方电网科学研究院有限责任公司 | Detection method and detection device for abnormal electricity consumption, electronic device and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112785056A (en) * | 2021-01-22 | 2021-05-11 | 杭州市电力设计院有限公司 | Short-term load prediction method based on fusion of Catboost and LSTM models |
CN112819136A (en) * | 2021-01-20 | 2021-05-18 | 南京邮电大学 | Time sequence prediction method and system based on CNN-LSTM neural network model and ARIMA model |
CN113159361A (en) * | 2020-12-03 | 2021-07-23 | 安徽大学 | Short-term load prediction method and system based on VDM and Stacking model fusion |
-
2021
- 2021-09-01 CN CN202111018225.6A patent/CN113705915A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113159361A (en) * | 2020-12-03 | 2021-07-23 | 安徽大学 | Short-term load prediction method and system based on VDM and Stacking model fusion |
CN112819136A (en) * | 2021-01-20 | 2021-05-18 | 南京邮电大学 | Time sequence prediction method and system based on CNN-LSTM neural network model and ARIMA model |
CN112785056A (en) * | 2021-01-22 | 2021-05-11 | 杭州市电力设计院有限公司 | Short-term load prediction method based on fusion of Catboost and LSTM models |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115146842A (en) * | 2022-06-24 | 2022-10-04 | 沈阳建筑大学 | Multivariate time series trend prediction method and system based on deep learning |
CN115329659A (en) * | 2022-07-18 | 2022-11-11 | 浙江大学 | Method and system for real-time early warning and intelligent control of dioxin emission in waste incinerator |
CN115329659B (en) * | 2022-07-18 | 2023-07-14 | 浙江大学 | Method and system for real-time early warning and intelligent control of dioxin emission of waste incinerator |
CN115860270A (en) * | 2023-02-21 | 2023-03-28 | 保定博堃元信息科技有限公司 | Network supply load prediction system and method based on LSTM neural network |
CN116227543A (en) * | 2023-05-09 | 2023-06-06 | 南方电网科学研究院有限责任公司 | Detection method and detection device for abnormal electricity consumption, electronic device and electronic equipment |
CN116227543B (en) * | 2023-05-09 | 2023-08-04 | 南方电网科学研究院有限责任公司 | Detection method and detection device for abnormal electricity consumption, electronic device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113705915A (en) | CNN-LSTM-ARIMA-based combined short-term power load prediction method | |
CN111191841B (en) | Power load prediction method and device, computer equipment and storage medium | |
CN109255505B (en) | Short-term load prediction method of multi-model fusion neural network | |
Fay et al. | On the influence of weather forecast errors in short-term load forecasting models | |
CN112016734A (en) | Stack type self-coding multi-model load prediction method and system based on LSTM | |
CN111695731B (en) | Load prediction method, system and equipment based on multi-source data and hybrid neural network | |
CN112819136A (en) | Time sequence prediction method and system based on CNN-LSTM neural network model and ARIMA model | |
CN108711847A (en) | A kind of short-term wind power forecast method based on coding and decoding shot and long term memory network | |
CN111027772A (en) | Multi-factor short-term load prediction method based on PCA-DBILSTM | |
CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
CN112633604A (en) | Short-term power consumption prediction method based on I-LSTM | |
Massaoudi et al. | Performance evaluation of deep recurrent neural networks architectures: Application to PV power forecasting | |
CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
CN116187555A (en) | Traffic flow prediction model construction method and prediction method based on self-adaptive dynamic diagram | |
CN114781744A (en) | Deep learning multi-step long radiance prediction method based on codec | |
CN112803398A (en) | Load prediction method and system based on empirical mode decomposition and deep neural network | |
CN114444561A (en) | PM2.5 prediction method based on CNNs-GRU fusion deep learning model | |
CN115049113A (en) | S2S power load prediction system based on time window and super-parameter adaptive selection | |
CN115688993A (en) | Short-term power load prediction method suitable for power distribution station area | |
CN116822920A (en) | Flow prediction method based on cyclic neural network | |
CN115238854A (en) | Short-term load prediction method based on TCN-LSTM-AM | |
CN114154753A (en) | Load prediction method and system | |
CN115860277B (en) | Data center energy consumption prediction method and system | |
CN116885699A (en) | Power load prediction method based on dual-attention mechanism | |
CN112580848A (en) | PT-LSTM-based time series prediction system and method |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20211126 |