CN111192158A - Transformer substation daily load curve similarity matching method based on deep learning - Google Patents
Transformer substation daily load curve similarity matching method based on deep learning Download PDFInfo
- Publication number
- CN111192158A CN111192158A CN201911265481.8A CN201911265481A CN111192158A CN 111192158 A CN111192158 A CN 111192158A CN 201911265481 A CN201911265481 A CN 201911265481A CN 111192158 A CN111192158 A CN 111192158A
- Authority
- CN
- China
- Prior art keywords
- daily load
- deep learning
- image
- transformer substation
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification 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
- G06N3/04—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
- G06N3/08—Learning methods
Abstract
The invention discloses a transformer substation daily load curve similarity matching method based on deep learning, which comprises the following steps of: s1, clustering daily load data of the transformer substation to obtain c categories of representative load data; s2, generating a load curve image according to the clustering result, and establishing a load image data set; s3 building a deep learning model based on the improved VGG16 and fully training on a data set; s4, inputting the data to be tested to the trained model for daily load curve type matching, and generating a visual document. Aiming at the defects of the prior art, the method introduces a deep learning model based on improved VGG16, starts from the aspect of image form, and can realize quick training and real-time matching on the premise of keeping the precision; the method comprises the steps that a single transformer substation is used as a research object, daily load data is used as research content, and load characteristics of the time scale of the transformer substation are deeply mined; and after the matching task is finished, generating a text document and an image document, and doubly visualizing the matching result.
Description
Technical Field
The invention relates to the field of load modeling of a power system, in particular to a transformer substation daily load curve similarity matching method based on deep learning.
Background
Load characteristics are important components of an electric power system, and have an important influence on analysis, design, and control of the electric power system as consumers of electric energy. Along with the implementation of the strategy of the energy internet, the load characteristics are more prominent, the energy internet requires certain intelligence in the generation, transmission, conversion and use of energy, and the research on the load characteristics is the key for promoting decision intelligence. Therefore, by utilizing an advanced algorithm, deep-level information of load characteristics is mined, and the method has practical significance for guiding power grid rolling planning, real-time scheduling and operation planning reliability evaluation. For the accuracy and practicality of load characteristic classification, many documents have conducted beneficial exploration.
The application of the improved adaptive fuzzy C-means algorithm in load characteristic classification (power system automation, 2018) such as Zhang Jianhua adopts the improved fuzzy C-means algorithm based on differential evolution adaptive optimization, and simulation experiments are carried out under different cluster numbers, and the accuracy is high. Patent document No. CN102999876A discloses a method for selecting a typical load characteristic substation, which takes a substation as a load characteristic as a research content, calculates the correlation coefficients of all substations in each load type by a gray correlation analysis method, and determines 10 typical sites. Patent document CN107194600A discloses a seasonal classification method for power load characteristics, which labels data according to a seasonal division method, and constructs a classification model by using a random forest algorithm to complete mechanical characteristic classification.
However, currently, a few studies on daily load characteristics are performed by taking a single substation as a unit, the algorithms classify daily load curves from a numerical similarity angle, and the studies on classifying the daily load curves from an image form angle by using deep learning are less.
Disclosure of Invention
The invention aims to overcome the technical defects, provides a transformer substation daily load curve similarity matching method based on deep learning, aims to provide a load characteristic matching method with high accuracy and strong real-time performance, introduces a deep learning model based on improved VGG16 for similarity matching, can realize real-time matching under the condition of ensuring precision, takes a single transformer substation as a research object, takes daily load data as research content, deeply excavates the load characteristic of a transformer substation time scale, generates a text document and an image document after completing a matching task, and visually matches results in multiple angles.
In order to achieve the object, the present invention adopts the following embodiments:
a transformer substation daily load curve similarity matching method based on deep learning comprises the following steps:
s1, clustering daily load data of the transformer substation to obtain c categories of representative load data;
s2, generating a load curve image according to the clustering result, and establishing a load image data set;
s3, building a deep learning model based on the improved VGG16 and fully training on a data set;
and S4, inputting the data to be tested to the trained model for daily load curve type matching, and generating a visual document.
The step S1 includes the following sub-steps:
s11, setting parameters of a fuzzy c-means algorithm (FCM), including category number c, fuzzy weight index f and an initial clustering center v, and clustering n daily load data of a certain year of the transformer substation;
s12, calculating a fuzzy membership matrix u and a center v of each category;
s13, calculating a fuzzy clustering target value, judging whether a condition is met, if so, terminating, if not, returning to the step S12, wherein the target function is as follows:
wherein J is a fuzzy clustering objective function, so that c characteristic categories can be obtained.
The step S2 includes the following sub-steps:
s21, obtaining daily load data of c categories according to the clustering result;
s22, visualizing the daily load data of each category to obtain a daily load curve image, establishing a daily load image database, adding category labels according to the category of each image, and keeping coordinate axis information of the curve image by using thicker lines;
and S23, dividing all images of the database into a training set, a verification set and a test according to the ratio of 7:2:1, wherein the training set is used for training a deep learning model, the verification set is used for adjusting parameters of the model, and the test set is used for verifying the feasibility of the model.
In step S21, the daily load data takes fifteen minutes as a node, and there are 96 nodes in total, and each node includes 330KV active power parameters.
The step S3 includes the following sub-steps:
s31, building a deep learning model based on the improved VGG16, wherein the model comprises thirteen convolutional layers, three fully-connected layers and a plurality of maxpool layers, the activation functions of the first fully-connected layer and the second fully-connected layer in the three fully-connected layers use tanh activation functions, and the formula is as follows:
s32, the third full connection layer in the three full connection layers adopts a softmax activation function, and the Loss function adopts a Focal local, wherein the softmax activation function is as follows:
the Focal local Loss function is as follows:
s33, inputting a load curve image generated according to a clustering result, establishing a load image data set to obtain a training set image, inputting the training set image to a deep learning model based on improved VGG16, performing iterative training on the whole model on the premise of freezing parameters of the first, second and third layers of convolutional layers, and setting iteration times;
s34, carrying out self-adaptive updating on the learning rate by using an Adam optimization algorithm, wherein the formula is as follows:
wherein the learning step η is 0.001, β1Is 0.9, β20.999, epsilon is 10-8;
And S35, finishing training based on the improved VGG16 model after finishing the set iteration number.
The step S4 includes the following sub-steps:
s41, inputting new daily load data to the trained model, and calculating the model in real time and completing similarity matching;
and S42, generating a visual document, wherein the visual document comprises a text document and an image document, the text document records the matching process and the result of the load characteristics, and the image document performs curvilinearized comparison on the daily load data and the matched characteristic types.
The invention has the beneficial effects that: 1) a deep learning model based on improved VGG16 is introduced, starting from the aspect of image form, and rapid training and real-time matching can be realized on the premise of maintaining precision;
2) the method comprises the steps that a single transformer substation is used as a research object, daily load data is used as research content, and load characteristics of the time scale of the transformer substation are deeply mined;
3) and generating a text document and an image document after the matching task is completed, and visualizing a matching result in multiple angles.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a daily load curve similarity matching architecture based on VGG 16.
Detailed Description
The structure and the beneficial effects of the invention are further explained in the following by combining the attached drawings.
As shown in fig. 1 and 2, a transformer substation daily load curve similarity matching method based on deep learning includes the following steps:
step one, clustering daily load data of a transformer substation to obtain representative load data of c categories;
the method comprises the following specific steps:
the first step is as follows: setting various parameters of a fuzzy c-means algorithm (FCM), including category number c, a fuzzy weight index f and an initial clustering center v, and clustering n daily load data of a transformer substation in a certain year;
the second step is that: and (3) calculating a fuzzy membership matrix u and the center v of each category, wherein the calculation formula is as follows:
the third step: calculating a fuzzy clustering target value, judging whether a condition is met, if so, stopping, and if not, returning to the second step, wherein the target function is as follows:
wherein J is a fuzzy clustering objective function, so that c characteristic categories can be obtained.
Step two: and generating a load curve image according to the clustering result, and establishing a load image data set.
The second step comprises the following specific steps:
the first step is as follows: and obtaining c categories of daily load data according to the clustering result, wherein the daily load data takes 15 minutes as a node, the total number of the nodes is 96, and each node comprises 330KV active power parameters.
The second step is that: visualizing the daily load data of each category to obtain a daily load curve image, establishing a daily load image database, adding category labels according to the category of each image, retaining coordinate axis information of the curve image, adopting thicker lines, and unifying the colors of the curves into black.
The third step: all images of the database are divided into a training set, a verification set and a test according to the proportion of 7:2:1, wherein the training set is used for training a deep learning model, the verification set is used for adjusting parameters of the model, and the test set is used for verifying the feasibility of the model.
Step three: a deep learning model based on the improved VGG16 was constructed and fully trained on the data set.
The third step comprises the following specific steps;
the first step is as follows: building a deep learning model based on the improved VGG16, wherein the model comprises 13 convolutional layers, 3 fully-connected layers and a plurality of maxpool layers, the first two fully-connected activation functions use tanh activation functions, and the formula is as follows:
the second step is that: the last full-link layer of the deep learning model adopts a softmax activation function, and the Loss function adopts a Focal local, wherein the softmax activation function is as follows:
the Focal local Loss function is as follows:
the third step: inputting the training set image obtained in the step of S2 to a deep learning model based on improved VGG16, and carrying out iterative training on the whole model on the premise of freezing the parameters of the previous 3 layers of convolutional layers, wherein the iteration times are set to be 500.
The fourth step: the learning rate is adaptively updated by using an Adam optimization algorithm, and the formula is as follows:
wherein the learning step η is 0.001, β1Is 0.9, β20.999, epsilon is 10-8。
The fifth step: a multi-stage transfer learning strategy is used in training the VGG16 network, the first 5 convolutional layer parameters are frozen in the first 400 iterations for training, and all layer parameters are fine-tuned in the last 100 iterations.
And a sixth step: and stopping network training after 500 iterations are completed or the loss value is less than 0.01, and completing training based on the improved VGG16 model, thereby completing training of the daily load curve similarity matching model based on deep learning.
Step four: and inputting the data to be tested to the trained model for daily load curve class matching, and generating a visual document.
The step 4 specifically comprises the following steps:
the first step is as follows: inputting new daily load data to the trained model, calculating the new daily load data in real time by the model and completing similarity matching
The second step is that: and generating a visual document, wherein the visual document comprises a text document and an image document, the text document records the matching process and the result of the load characteristics, and the image document performs curvilinearization comparison on the daily load data and the matched characteristic category.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (6)
1. A transformer substation daily load curve similarity matching method based on deep learning is characterized by comprising the following steps:
s1, clustering daily load data of the transformer substation to obtain c categories of representative load data;
s2, generating a load curve image according to the clustering result, and establishing a load image data set;
s3, building a deep learning model based on the improved VGG16 and fully training on a data set;
and S4, inputting the data to be tested to the trained model for daily load curve type matching, and generating a visual document.
2. The transformer substation daily load curve similarity matching method based on deep learning according to claim 1, characterized in that: the step S1 includes the following sub-steps:
s11, setting parameters of a fuzzy c-means algorithm (FCM), including category number c, fuzzy weight index f and an initial clustering center v, and clustering n daily load data of a certain year of the transformer substation;
s12, calculating a fuzzy membership matrix u and a center v of each category;
s13, calculating a fuzzy clustering target value, judging whether a condition is met, if so, terminating, if not, returning to the step S12, wherein the target function is as follows:
wherein J is a fuzzy clustering objective function, so that c characteristic categories can be obtained.
3. The transformer substation daily load curve similarity matching method based on deep learning according to claim 1, characterized in that: the step S2 includes the following sub-steps:
s21, obtaining daily load data of c categories according to the clustering result;
s22, visualizing the daily load data of each category to obtain a daily load curve image, establishing a daily load image database, adding category labels according to the category of each image, and keeping coordinate axis information of the curve image by using thicker lines;
and S23, dividing all images of the database into a training set, a verification set and a test according to the ratio of 7:2:1, wherein the training set is used for training a deep learning model, the verification set is used for adjusting parameters of the model, and the test set is used for verifying the feasibility of the model.
4. The transformer substation daily load curve similarity matching method based on deep learning according to claim 1, characterized in that: in step S21, the daily load data takes fifteen minutes as a node, and there are 96 nodes in total, and each node includes 330KV active power parameters.
5. The transformer substation daily load curve similarity matching method based on deep learning according to claim 1, characterized in that: the step S3 includes the following sub-steps:
s31, building a deep learning model based on the improved VGG16, wherein the model comprises thirteen convolutional layers, three fully-connected layers and a plurality of maxpool layers, the activation functions of the first fully-connected layer and the second fully-connected layer in the three fully-connected layers use tanh activation functions, and the formula is as follows:
s32, the third full connection layer in the three full connection layers adopts a softmax activation function, and the Loss function adopts a Focal local, wherein the softmax activation function is as follows:
the Focal local Loss function is as follows:
s33, inputting a load curve image generated according to a clustering result, establishing a load image data set to obtain a training set image, inputting the training set image to a deep learning model based on improved VGG16, performing iterative training on the whole model on the premise of freezing parameters of the first, second and third layers of convolutional layers, and setting iteration times;
s34, carrying out self-adaptive updating on the learning rate by using an Adam optimization algorithm, wherein the formula is as follows:
wherein the learning step η is 0.001, β1Is 0.9, β20.999, epsilon is 10-8;
And S35, finishing training based on the improved VGG16 model after finishing the set iteration number.
6. The transformer substation daily load curve similarity matching method based on deep learning according to claim 1, characterized in that: the step S4 includes the following sub-steps:
s41, inputting new daily load data to the trained model, and calculating the model in real time and completing similarity matching;
and S42, generating a visual document, wherein the visual document comprises a text document and an image document, the text document records the matching process and the result of the load characteristics, and the image document performs curvilinearized comparison on the daily load data and the matched characteristic types.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911265481.8A CN111192158A (en) | 2019-12-11 | 2019-12-11 | Transformer substation daily load curve similarity matching method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911265481.8A CN111192158A (en) | 2019-12-11 | 2019-12-11 | Transformer substation daily load curve similarity matching method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111192158A true CN111192158A (en) | 2020-05-22 |
Family
ID=70710971
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911265481.8A Pending CN111192158A (en) | 2019-12-11 | 2019-12-11 | Transformer substation daily load curve similarity matching method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111192158A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112418331A (en) * | 2020-11-26 | 2021-02-26 | 国网甘肃省电力公司电力科学研究院 | Clustering fusion-based semi-supervised learning pseudo label assignment method |
CN114254146A (en) * | 2020-09-21 | 2022-03-29 | 京东方科技集团股份有限公司 | Image data classification method, device and system |
CN114612744A (en) * | 2022-03-10 | 2022-06-10 | 平安科技(深圳)有限公司 | Detection model training method, vehicle damage detection method and terminal equipment |
-
2019
- 2019-12-11 CN CN201911265481.8A patent/CN111192158A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114254146A (en) * | 2020-09-21 | 2022-03-29 | 京东方科技集团股份有限公司 | Image data classification method, device and system |
CN112418331A (en) * | 2020-11-26 | 2021-02-26 | 国网甘肃省电力公司电力科学研究院 | Clustering fusion-based semi-supervised learning pseudo label assignment method |
CN114612744A (en) * | 2022-03-10 | 2022-06-10 | 平安科技(深圳)有限公司 | Detection model training method, vehicle damage detection method and terminal equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jiang et al. | Scenario generation for wind power using improved generative adversarial networks | |
Wasanapradit et al. | Solving mixed-integer nonlinear programming problems using improved genetic algorithms | |
CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
CN109345027B (en) | Micro-grid short-term load prediction method based on independent component analysis and support vector machine | |
CN110212528B (en) | Power distribution network measurement data missing reconstruction method | |
CN109242223B (en) | Quantum support vector machine evaluation and prediction method for urban public building fire risk | |
CN111192158A (en) | Transformer substation daily load curve similarity matching method based on deep learning | |
CN110942205A (en) | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM | |
CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
CN105335619A (en) | Collaborative optimization method applicable to parameter back analysis of high calculation cost numerical calculation model | |
CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
Yan et al. | Trustworthiness evaluation and retrieval-based revision method for case-based reasoning classifiers | |
Wang et al. | Short-term load forecasting with LSTM based ensemble learning | |
CN112163671A (en) | New energy scene generation method and system | |
Zeng et al. | Short-term load forecasting of smart grid systems by combination of general regression neural network and least squares-support vector machine algorithm optimized by harmony search algorithm method | |
CN114282646B (en) | Optical power prediction method and system based on two-stage feature extraction and BiLSTM improvement | |
CN115358437A (en) | Power supply load prediction method based on convolutional neural network | |
CN112749517A (en) | Tolerance optimization distribution method | |
CN117113086A (en) | Energy storage unit load prediction method, system, electronic equipment and medium | |
CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
CN116523001A (en) | Method, device and computer equipment for constructing weak line identification model of power grid | |
CN111353525A (en) | Modeling and missing value filling method for unbalanced incomplete data set | |
CN116843057A (en) | Wind power ultra-short-term prediction method based on LSTM-ViT | |
CN116316629A (en) | Data-driven optimal power flow calculation method considering topology feature learning | |
CN116303386A (en) | Intelligent interpolation method and system for missing data based on relational graph |
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 |