CN114818871A - Abnormal electricity utilization detection method for power distribution network with distributed power supply - Google Patents
Abnormal electricity utilization detection method for power distribution network with distributed power supply Download PDFInfo
- Publication number
- CN114818871A CN114818871A CN202210327138.7A CN202210327138A CN114818871A CN 114818871 A CN114818871 A CN 114818871A CN 202210327138 A CN202210327138 A CN 202210327138A CN 114818871 A CN114818871 A CN 114818871A
- Authority
- CN
- China
- Prior art keywords
- data
- abnormal
- features
- load
- detection method
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 61
- 238000001514 detection method Methods 0.000 title claims abstract description 57
- 230000005611 electricity Effects 0.000 title claims abstract description 39
- 238000000605 extraction Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000012706 support-vector machine Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 238000011176 pooling Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000013527 convolutional neural network Methods 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 238000012966 insertion method Methods 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 15
- 230000008859 change Effects 0.000 abstract description 11
- 230000006399 behavior Effects 0.000 description 8
- 230000009286 beneficial effect Effects 0.000 description 5
- 238000010248 power generation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an abnormal power utilization detection method for a power distribution network containing distributed power supplies, which comprises the following steps: data acquisition and data preprocessing; constructing an input feature set and converting the input feature set into a pixel picture; establishing a convolution neural network, a gate control circulation unit and a support vector machine combined abnormal electricity utilization characteristic extraction and detection model; and training the detection model by adopting the training set data. Compared with the traditional method that only historical load data and temperature meteorological data are considered by the input feature set, the method has the advantages that the influence factors of the net load containing the distributed power supply can be comprehensively reflected by the input feature set, the change mode of the net load can be analyzed, and the accuracy of abnormal electricity utilization detection can be improved. The invention fully excavates the local characteristics, the global characteristics and the time sequence change rule of the net load, reveals the condition that the net load is influenced by meteorological factors, further comprehensively grasps the change mode of the net load and realizes the abnormal electricity detection with higher precision.
Description
Technical Field
The invention relates to the technical field of power grid power utilization abnormity detection, in particular to an abnormal power utilization detection method for a power distribution network with distributed power supplies.
Background
Abnormal electricity consumption behaviors such as electricity stealing and the like are important reasons for power transmission and distribution loss of a power grid, can cause great economic loss of power grid enterprises, and are important problems expected to be solved by power departments of various countries. The accurate and efficient abnormal electricity utilization detection method can quickly and accurately find electricity stealing behaviors, reduces economic loss caused by electricity stealing, and has important value.
The existing abnormal electricity detection methods can be classified into a method based on a system state and a method based on data driving. The method based on the system state utilizes the voltage, current and node power equivalent measurement data of the power distribution network to carry out state estimation, and then abnormal power utilization users are detected. The method needs complete network topology structure parameters, has high cost and is difficult to implement in practice.
The data-driven abnormal electricity utilization detection method is directly based on the characteristics of the shape, the electricity distribution and the like of a user electricity utilization curve, realizes the identification of abnormal electricity utilization users by means of a clustering technology and a classification technology, has the advantages of low cost, high efficiency, easiness in implementation and the like, and is a current research hotspot.
The existing abnormal electricity utilization detection method based on data driving generally comprises three links:
(1) constructing an input feature set, namely determining input data of a detection model;
(2) extracting features, namely further extracting important features beneficial to anomaly detection from input data;
(3) and (4) abnormal identification, namely identifying normal and abnormal electricity utilization behaviors according to the extracted important features.
When the method is oriented to a power distribution network with a distributed power supply, the existing abnormal power utilization detection method has the following defects in the above links:
in the link of constructing an input feature set, the existing method does not construct the input features of an anomaly detection model aiming at the characteristics of loads containing distributed power supplies. For example, the patent "a power consumption anomaly detection method and system based on a deep neural network" (CN202111237405.3) adopts historical power consumption data and environmental temperature data as input of an anomaly detection model, and the patent "a power distribution network loss analysis method and system based on multi-source measurement data" (CN202111433526.5) adopts power consumption data, temperature climate data and station area information as input data of the detection model. However, using these data alone as an input feature set does not reflect the characteristics of the load containing the distributed power supply. This is because, in the case of accessing a distributed power supply, the load is affected by the coupling of multiple factors including the user's electricity usage habits, the rated capacity of the fan or photovoltaic panel, the wind speed, irradiance, temperature, and the like. The conventional abnormal electricity utilization detection method does not take the factors into full consideration, so that the change mode of the load cannot be accurately excavated, and the method cannot be applied to abnormal electricity utilization detection for a power distribution network with a distributed power supply.
In the step of feature extraction, the feature extraction model of the existing abnormal electricity utilization detection method is only suitable for conventional loads, and a feature extraction model specially aiming at a distributed power supply is not provided. Therefore, when the conventional abnormal electricity utilization detection method is applied to a power distribution network with a distributed power supply, the detection precision is low.
With the access of a large amount of distributed renewable energy sources on the load side of the current power distribution network, the power load becomes the net load obtained by subtracting the generated power of the distributed power sources from the actual power consumption. The net load contains the actual electricity utilization component of the user and the renewable energy power generation component. This means that the electrical load is not only affected by the electrical behavior of the user, but also by multiple meteorological factors such as wind speed, irradiance, temperature, etc., and the load characteristics become more complex and more uncertain. The existing abnormal electricity utilization detection method based on data driving mainly aims at conventional loads with strong regularity and is difficult to be suitable for net loads with more complex characteristics and distributed power supplies. Therefore, it is necessary to develop an abnormal electricity usage detection method for a power distribution network including distributed power sources.
Disclosure of Invention
The invention aims to provide an abnormal power utilization detection method for a power distribution network with distributed power supplies, which can solve the problem that the detection precision is low when the abnormal power utilization detection method in the prior art is used for the power distribution network with the distributed power supplies.
The purpose of the invention is realized by the following technical scheme:
an abnormal power utilization detection method for a power distribution network with distributed power supplies comprises the following steps:
step S1, data acquisition and data preprocessing;
step S2, constructing an input feature set and converting the input feature set into a pixel picture;
step S3, establishing a convolution neural network CNN, a gating circulation unit GRU, a support vector machine SVM and a combined abnormal electricity utilization characteristic extraction and detection model CNN-GRU-SVM;
and S4, training the detection model CNN-GRU-SVM described in the step S3 by adopting training set data.
Further, the step S1 includes:
s101, acquiring historical payload data, historical meteorological data, parameter data of installed distributed power supplies and user tag data of a plurality of users in an area to be detected;
s102, filling and replacing missing data and abnormal mutation data by adopting a mean value insertion method; for the repeated data, the in-out deletion is processed;
step S103, preprocessing the data according to the following steps of 8: 2 into training and test sets.
Further, the configuration input feature set comprises configuration power load features, configuration new energy related weather features, configuration new energy equipment features and configuration date features.
Further, the power load characteristics are divided into load curve image characteristics and load statistical characteristics, and the load curve image characteristics are obtained by converting the user net load matrix into a pixel picture.
Further, the step S3 includes:
step S301, for load curve image characteristics in the power load characteristics, performing characteristic extraction by adopting the combination of n convolutional layers and pooling layers to obtain deep layer characteristics F1;
step S302, for load statistical characteristics in the power load characteristics, performing characteristic extraction by adopting k GRU networks to obtain deep characteristics F2;
step S303, for the related meteorological features of the new energy, performing feature extraction by adopting the combination of m convolutional layers and pooling layers to obtain deep features F3;
step S304, extracting input features by adopting a fully-connected Dense network for the features of the new energy equipment to obtain deep features F4;
s305, extracting input features to obtain deep features F5 by adopting a fully-connected Dense network for date features;
s306, inputting the deep feature F1, the deep feature F2 and the deep feature F3 into a Dense network, and performing further feature extraction to obtain a new deep feature F6;
step S307, inputting the deep features F4, F5 and F6 into a Dense network for further feature extraction to obtain new deep features F7;
and S308, inputting the obtained new deep features F6 and F7 into an SVM classifier to obtain an identification result of the user power consumption behavior.
Further, the calculation formula of the convolutional layer in step S301 is:
c i ＝f c (A i *w c,i +b c,i )；
wherein, c i The output of the ith convolution layer, i is 1,2, …, n; a. the i 、w c,i 、b c,i The input, weight and bias of the ith convolution layer respectively; f. of c Is a convolutional layer activation function; and the dot product calculation is represented, namely the dot product calculation is performed after the multiplication of the corresponding elements of the two matrixes.
Further, the calculation formula of the pooling layer in step S301 is as follows:
d i ＝w d,i *z i ；
wherein d is i Is the output of the ith pooling layer; w is a d,i The weight value of the ith pooling layer; z is a radical of i Is d i A corresponding pooling zone.
Further, the calculation formula of the GRU network in step S302 is as follows:
wherein r is t 、o t 、
h t Outputs of a reset gate, an update gate, a candidate gate and a state gate which are respectively input in the t step in the GRU unit; w r 、W o 、
Respectively resetting the weight of the gate, updating the weight of the gate and selecting the weight of the gate; x is the number of t Input data for GRU; y is t Output data of GRU; σ is the activation function.
Further, in step S4, inputting the test set or the data set to be detected into the trained detection model CNN-GRU-SVM to obtain an identification result of abnormal power consumption; and evaluating the identification result by adopting the precision index and the recall index.
Further, the precision ratio is defined as:
the recall ratio is defined as:
wherein, TP is the number of samples identified as normal user and normal user; FP represents the number of samples that are identified as normal users but are abnormal users; FN represents the number of samples identified as abnormal users but actually normal users.
The invention constructs an input feature set comprising a power load feature, a new energy related meteorological feature, a new energy device feature and a date feature. Particularly, the power load characteristic and the new energy related meteorological characteristic are constructed in a pixel picture mode, and the CNN network is more beneficial to extracting potential electricity utilization information. Compared with the traditional method that only historical load data and temperature meteorological data are considered by the input feature set, the input feature set can comprehensively reflect the influence factors of the net load containing the distributed power supply, is more beneficial to analyzing the change mode of the net load, and is further beneficial to improving the accuracy of abnormal electricity detection.
The invention constructs a model (CNN-GRU-SVM) for extracting and detecting abnormal electricity utilization characteristics aiming at the characteristics of the net load containing the distributed power supply. Particularly, deep feature extraction is carried out on load curve image features and new energy related meteorological features by adopting a multilayer CNN network, and local and global change rules of net load and meteorological are deeply excavated; deep feature extraction is carried out on the load statistical features by adopting a GRU network, and the time sequence change rule of the net load is mined; the extracted payload characteristics and meteorological characteristics are fused by adopting a Dense network, so that the correlation between the payload and the meteorological is favorably mined, and the detection precision is improved; and further fusing the extracted deep level features by adopting a Dense network, and then inputting the fused deep level features into an SVM classifier to realize high-precision abnormal electricity utilization identification.
The CNN-GRU-SVM model provided by the invention can fully mine the local characteristics, the global characteristics and the time sequence change rule of the net load, and reveal the condition that the net load is influenced by meteorological factors, thereby more comprehensively grasping the change mode of the net load and realizing the abnormal electricity utilization detection with higher precision.
Drawings
FIG. 1 is a schematic diagram of a pixel picture converted from a user payload matrix according to the present invention;
FIG. 2 is a schematic view of a pixel picture transformed from a matrix formed by meteorological data according to the present invention;
FIG. 3 is a schematic diagram of a detection model CNN-GRU-SVM of the present invention.
Detailed Description
The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
The embodiments of the present disclosure are described below with specific examples, and other advantages and effects of the present disclosure will be readily apparent to those skilled in the art from the disclosure of the present disclosure. It is to be understood that the described embodiments are merely illustrative of some, and not restrictive, of the embodiments of the disclosure. The disclosure may be embodied or carried out in various other specific embodiments, and various modifications and changes may be made in the details within the description without departing from the spirit of the disclosure. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
The invention relates to an abnormal power utilization detection method for a power distribution network containing distributed power supplies, which comprises the following steps of:
and step S1, data acquisition and data preprocessing.
Specifically, step S1 includes:
step S101, obtaining historical net load data, historical meteorological data, parameter data of installed distributed power supplies and user tag data of a plurality of users in an area to be detected.
S102, filling and replacing missing data and abnormal mutation data by adopting a mean value insertion method; and for repeated data, performing in-out deletion processing.
Step S103, preprocessing the data according to the following steps of 8: 2 into training and test sets.
And step S2, constructing an input feature set and converting the input feature set into a pixel picture.
Further, the configuration input feature set comprises configuration power load features, configuration new energy related weather features, configuration new energy equipment features and configuration date features. The method comprises the following specific steps:
further, the power load characteristics are divided into load curve image characteristics and load statistical characteristics: the load curve image characteristics are obtained by converting a user net load matrix into a pixel picture. Each column in the user payload matrix corresponds to energy consumption at a certain time of day and each row corresponds to a different day. For example, the user's 2-week electrical net load data forms a 14 x 24 matrix, where 14 is the number of days and 24 is the load at 24 points per day. After converting the matrix into a pixel picture, the picture as shown in fig. 1 can be obtained.
The advantages of using a pixel picture to represent the load curve characteristics are: after the load characteristics are converted into the pixel picture, the line, the shape and the color of the image contain the user electricity consumption behavior information, and then the user electricity consumption information contained in the picture can be captured by using a Convolutional Neural Network (CNN) with a translation invariant characteristic. Compared with a numerical matrix, CNN is better at extracting information from a picture, and therefore load data is converted into a pixel image, which is more beneficial to extraction of load characteristics.
Further, the load statistical characteristics include: daily electricity consumption, daily maximum load, daily minimum load, daily average load. The load statistical characteristics reflect the distribution of the user net load.
Further, the new energy related meteorological features comprise meteorological data such as wind speed, solar irradiance, temperature, humidity and the like corresponding to the electric load characteristic time. And converting the matrix formed by the meteorological data into a pixel picture. Taking the solar irradiance data as an example, the pixel image is converted into a pixel image as shown in fig. 2.
Further, the new energy device characteristics include maximum capacity, rated temperature, fan cut-out wind speed, and the like of the distributed power supply device. The new energy equipment parameters are closely related to the power generation power of the distributed power supply, and the new energy equipment parameters are used as input characteristic data, so that the change rule of the net load can be analyzed, and the accuracy of abnormal power utilization detection is improved.
Further, the date characteristics include: date, week, month, quarter information. The characteristics of the power utilization behavior of the user and the characteristics of the distributed power generation are different at different dates, weeks, months and seasons, and the date characteristic information is used as the input characteristic, so that the understanding of the change rule of the net load is facilitated, and the accuracy of abnormal power utilization detection is further improved.
Step S3, establishing a model (CNN-GRU-SVM) for extracting and detecting abnormal electricity consumption characteristics, which is formed by combining a Convolutional Neural Network (CNN), a Gated current Unit (GRU) and a Support Vector Machine (SVM).
The established CNN-GRU-SVM model is shown in fig. 3, and step S3 includes:
step S301, for the load curve image features in the power load features, feature extraction is performed by using a combination of n convolutional layers and pooling layers, and a deep feature extracted from the load curve image by CNN is denoted as F1.
Wherein, the calculation formula of the convolution layer is as follows:
c i ＝f c (A i *w c,i +b c,i )(1)
(1) in the formula, c i Is the output of the ith convolutional layer, i ═ 1,2, …, n; a. the i 、w c,i 、b c,i The input, weight and bias of the ith convolution layer respectively; f. of c Is a convolutional layer activation function; and the dot product calculation is represented, namely the dot product calculation is performed after the multiplication of the corresponding elements of the two matrixes.
The formula of the pooling layer is as follows:
d i ＝w d,i *z i (2)
(2) in the formula (d) i Is the output of the ith pooling layer; w is a d,i Is the ithThe weight of each pooling layer; z is a radical of i Is d i A corresponding pooling zone.
Step S302, for the load statistical characteristics in the power load characteristics, performing characteristic extraction by adopting k GRU networks, and recording deep characteristics obtained by GRU extraction from the load statistical characteristics as F2.
Further, the calculation formula of the GRU network is as follows:
(3) in the formula, r t 、o t 、
h t Outputs of a reset gate, an update gate, a candidate gate and a state gate which are respectively input in the t step in the GRU unit; w r 、W o 、
Respectively resetting the weight of the gate, updating the weight of the gate and selecting the weight of the gate; x is the number of t Input data for GRU; y is t Output data of GRU; σ is the activation function.
And step S303, for the related meteorological features of the new energy, performing feature extraction by adopting the combination of the m convolutional layers and the pooling layers.
And recording the deep features extracted by the CNN from the related meteorological features of the new energy as F3.
And S304, extracting input characteristics of the new energy equipment by adopting a fully-connected Dense network.
And (5) recording deep features extracted from the features of the new energy equipment by the Dense network as F4.
And S305, extracting the input features by adopting a fully-connected Dense network for the date features.
And F5, marking the deep features extracted from the date feature features by the Dense network as F5.
And S306, inputting the deep features F1, F2 and F3 into a Dense network, and performing further feature extraction to obtain a new deep feature, which is marked as F6.
And step S307, inputting the deep features F4, F5 and F6 into a Dense network, and performing further feature extraction to obtain a new deep feature, which is marked as F7.
And S308, inputting the extracted comprehensive characteristics F6 and F7 into an SVM classifier to obtain an identification result of the power utilization behavior of the user.
And S4, training the CNN-GRU-SVM model in the step S3 by adopting training set data. And (3) optimally selecting the hyper-parameters n, k and m mentioned in the step (3) and the number of neurons in hidden layers of the CNN network, the GRU network and the Dense network by adopting a grid search method.
And inputting the test set or the data set to be detected into the trained model to obtain the identification result of the abnormal power utilization. And evaluating the identification result by adopting the precision index and the recall index.
The precision ratio is defined as:
recall is defined as:
wherein, TP is the number of samples identified as normal user and normal user; FP represents the number of samples that are identified as normal users but are abnormal users; FN represents the number of samples identified as abnormal users but actually normal users.
Taking a certain payload data set as an example, simulation analysis is performed, wherein the total number of samples is 4619, the number of normal users is 2772, and the number of abnormal users is 1874. The abnormal electricity utilization detection is carried out by adopting the method of the invention, and the results are shown in the following table 1:
TABLE 1 detection accuracy of the method of the present invention for a certain example
Wherein the average precision ratio is 0.87, and the average recall ratio is 0.83. Therefore, the method provided by the invention has higher abnormal electricity utilization detection precision aiming at the net load containing the distributed power supply.
In the present invention, unless otherwise expressly stated or limited, the first feature may be "on" the second feature in direct contact with the second feature, or the first and second features may be in indirect contact via an intermediate. "plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "connected," "secured," and the like are to be construed broadly and can, for example, be fixedly connected, detachably connected, or integrally formed; may be mechanically coupled, may be electrically coupled or may be in communication with each other; they may be directly connected or indirectly connected through intervening media, or they may be connected internally or in any other suitable relationship, unless expressly stated otherwise. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
The above description is for the purpose of illustrating embodiments of the invention and is not intended to limit the invention, and it will be apparent to those skilled in the art that any modification, equivalent replacement, or improvement made without departing from the spirit and principle of the invention shall fall within the protection scope of the invention.
Claims (10)
1. An abnormal electricity utilization detection method for a power distribution network with distributed power supplies is characterized by comprising the following steps:
step S1, data acquisition and data preprocessing;
step S2, constructing an input feature set and converting the input feature set into a pixel picture;
step S3, establishing a convolution neural network CNN, a gating cycle unit GRU, a support vector machine SVM and a combined abnormal electricity utilization characteristic extraction and detection model CNN-GRU-SVM;
and S4, training the detection model CNN-GRU-SVM described in the step S3 by adopting training set data.
2. The abnormal power consumption detection method for the power distribution network with the distributed power supplies according to claim 1, wherein the step S1 includes:
s101, acquiring historical payload data, historical meteorological data, parameter data of installed distributed power supplies and user tag data of a plurality of users in an area to be detected;
s102, filling and replacing missing data and abnormal mutation data by adopting a mean value insertion method; for the repeated data, the in-out deletion is processed;
step S103, preprocessing the data according to the following steps of 8: 2 into training and test sets.
3. The abnormal power utilization detection method for the power distribution network with the distributed power supplies according to claim 1, wherein the configuration input feature set comprises configuration power load features, configuration new energy related meteorological features, configuration new energy equipment features and configuration date features.
4. The abnormal power consumption detection method for the power distribution network with the distributed power supplies according to claim 3, wherein the power load characteristics are divided into load curve image characteristics and load statistical characteristics, and the load curve image characteristics are obtained by converting a user net load matrix into a pixel picture.
5. The abnormal power consumption detection method for the power distribution network with the distributed power supplies according to claim 1, wherein the step S3 includes:
step S301, for load curve image characteristics in the power load characteristics, performing characteristic extraction by adopting the combination of n convolutional layers and pooling layers to obtain deep layer characteristics F1;
step S302, for load statistical characteristics in the power load characteristics, performing characteristic extraction by adopting k GRU networks to obtain deep characteristics F2;
step S303, for the related meteorological features of the new energy, performing feature extraction by adopting the combination of m convolutional layers and pooling layers to obtain deep features F3;
step S304, extracting input features by adopting a fully-connected Dense network for the features of the new energy equipment to obtain deep features F4;
s305, extracting input features to obtain deep features F5 by adopting a fully-connected Dense network for date features;
s306, inputting the deep feature F1, the deep feature F2 and the deep feature F3 into a Dense network, and performing further feature extraction to obtain a new deep feature F6;
step S307, inputting the deep feature F4, the deep feature F5 and the deep feature F6 into a Dense network, and performing further feature extraction to obtain a new deep feature F7;
and S308, inputting the obtained new deep features F6 and F7 into an SVM classifier to obtain an identification result of the user power consumption behavior.
6. The abnormal power consumption detection method for the power distribution network with the distributed power supplies as claimed in claim 5, wherein the calculation formula of the convolution layer in the step S301 is as follows:
c i ＝f c (A i *w c,i +b c,i )；
wherein, c i Is the output of the ith convolutional layer, i ═ 1,2, …, n; a. the i 、w c,i 、b c,i The input, weight and bias of the ith convolution layer respectively; f. of c Is a convolutional layer activation function; and the dot product calculation is represented, namely the dot product calculation is performed after the multiplication of the corresponding elements of the two matrixes.
7. The abnormal power consumption detection method for the power distribution network with the distributed power supplies as claimed in claim 5, wherein the calculation formula of the pooling layer in the step S301 is as follows:
d i ＝w d,i *z i ；
wherein d is i Is the output of the ith pooling layer; w is a d,i The weight value of the ith pooling layer; z is a radical of i Is d i A corresponding pooling zone.
8. The abnormal power consumption detection method for the power distribution network with the distributed power supplies as claimed in claim 5, wherein the calculation formula of the GRU network in the step S302 is as follows:
wherein r is t 、o t 、
h t Outputs of a reset gate, an update gate, a candidate gate and a state gate which are respectively input in the t step in the GRU unit; w r 、W o 、
Respectively resetting the weight of the gate, updating the weight of the gate and selecting the weight of the gate; x is the number of t Input data for GRU; y is t Output data of GRU; σ is the activation function.
9. The abnormal power consumption detection method for the power distribution network with the distributed power supplies as claimed in claim 1, wherein in step S4, the test set or the data set to be detected is input into a trained detection model CNN-GRU-SVM to obtain an identification result of the abnormal power consumption; and evaluating the identification result by adopting the precision index and the recall index.
10. The abnormal power consumption detection method for the power distribution network with the distributed power supplies according to claim 9, wherein the precision ratio is defined as:
the recall ratio is defined as:
wherein, TP is the number of samples identified as normal user and normal user; FP represents the number of samples that are identified as normal users but are abnormal users; FN represents the number of samples identified as abnormal users but actually normal users.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210327138.7A CN114818871A (en) | 2022-03-30 | 2022-03-30 | Abnormal electricity utilization detection method for power distribution network with distributed power supply |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210327138.7A CN114818871A (en) | 2022-03-30 | 2022-03-30 | Abnormal electricity utilization detection method for power distribution network with distributed power supply |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114818871A true CN114818871A (en) | 2022-07-29 |
Family
ID=82533633
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210327138.7A Pending CN114818871A (en) | 2022-03-30 | 2022-03-30 | Abnormal electricity utilization detection method for power distribution network with distributed power supply |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114818871A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115409296A (en) * | 2022-11-02 | 2022-11-29 | 国网江西省电力有限公司电力科学研究院 | Method for predicting net load probability of active power distribution network |
-
2022
- 2022-03-30 CN CN202210327138.7A patent/CN114818871A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115409296A (en) * | 2022-11-02 | 2022-11-29 | 国网江西省电力有限公司电力科学研究院 | Method for predicting net load probability of active power distribution network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647716B (en) | Photovoltaic array fault diagnosis method based on composite information | |
| Chen et al. | Parameters identification of photovoltaic models using hybrid adaptive Nelder-Mead simplex algorithm based on eagle strategy | |
| Rao et al. | Machine learning for solar array monitoring, optimization, and control | |
| Liu et al. | Fault diagnosis approach for photovoltaic array based on the stacked auto-encoder and clustering with IV curves | |
| Feng et al. | A taxonomical review on recent artificial intelligence applications to PV integration into power grids | |
| CN101728868B (en) | Method for classification and forecast of remote measuring power load patterns | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| Liu et al. | Intelligent fault diagnosis of photovoltaic array based on variable predictive models and I–V curves | |
| CN111859283B (en) | Scene generation method considering multi-energy charge time sequence and correlation | |
| CN111461921A (en) | Load modeling typical user database updating method based on machine learning | |
| CN112949207A (en) | Short-term load prediction method based on improved least square support vector machine | |
| CN114818871A (en) | Abnormal electricity utilization detection method for power distribution network with distributed power supply | |
| Huang et al. | Parameter identification for photovoltaic models using an improved learning search algorithm | |
| Lai et al. | Sub-region division based short-term regional distributed PV power forecasting method considering spatio-temporal correlations | |
| Jønler et al. | Probabilistic forecasts of global horizontal irradiance for solar systems | |
| CN112508278A (en) | Multi-connected system load prediction method based on evidence regression multi-model | |
| CN117439045A (en) | Multi-element load prediction method for comprehensive energy system | |
| CN116662860A (en) | User portrait and classification method based on energy big data | |
| Rosato et al. | 2-D convolutional deep neural network for multivariate energy time series prediction | |
| Adhya et al. | Stacking ensemble based fault diagnosis approach for improved operation of photovoltaic arrays | |
| CN116340534A (en) | Knowledge graph construction method and system for identifying new energy abnormal data | |
| CN113537607B (en) | Power failure prediction method | |
| CN114595884A (en) | Genetic intelligent optimization neural network wind power generation equipment temperature prediction method | |
| Rosato et al. | A Combined Deep Learning Approach for Time Series Prediction in Energy Environments | |
| CN110991083B (en) | Photovoltaic power station model determining method, device, equipment and storage medium |
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 |