CN118035646A - Power grid fault threat degree assessment method, device, equipment and storage medium - Google Patents
Power grid fault threat degree assessment method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN118035646A CN118035646A CN202410260246.6A CN202410260246A CN118035646A CN 118035646 A CN118035646 A CN 118035646A CN 202410260246 A CN202410260246 A CN 202410260246A CN 118035646 A CN118035646 A CN 118035646A
- Authority
- CN
- China
- Prior art keywords
- fault
- threat
- power
- feature
- power utilization
- 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 54
- 238000005065 mining Methods 0.000 claims abstract description 53
- 239000013598 vector Substances 0.000 claims abstract description 52
- 238000013210 evaluation model Methods 0.000 claims abstract description 24
- 230000005611 electricity Effects 0.000 claims description 70
- 239000011159 matrix material Substances 0.000 claims description 57
- 238000011156 evaluation Methods 0.000 claims description 56
- 238000000605 extraction Methods 0.000 claims description 40
- 230000006399 behavior Effects 0.000 claims description 36
- 238000011176 pooling Methods 0.000 claims description 26
- 238000013528 artificial neural network Methods 0.000 claims description 23
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 11
- 125000004122 cyclic group Chemical group 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 230000000306 recurrent effect Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 29
- 239000000284 extract Substances 0.000 abstract description 11
- 238000001514 detection method Methods 0.000 description 12
- 230000008859 change Effects 0.000 description 10
- 238000012549 training Methods 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000002123 temporal effect Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000007477 logistic regression Methods 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 4
- 230000007704 transition Effects 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000012854 evaluation process Methods 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 101100001676 Emericella variicolor andK gene Proteins 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a power grid fault threat degree assessment method, a device, equipment and a storage medium, wherein the method comprises the following steps: acquiring power utilization time sequence data of a power grid user in a preset period, and converting the power utilization time sequence data into corresponding power utilization image data; acquiring space-time feature vectors corresponding to the power utilization image data according to a preset feature mining model, and identifying a plurality of power utilization fault types to be evaluated according to the space-time feature vectors; determining the relative priority corresponding to each power failure type according to a preset failure threat degree evaluation model; the relative priority is used for indicating the priority of the power consumption fault type relative to other power consumption fault types; and evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority. The method can efficiently and accurately extract key power fault characteristics, comprehensively and comprehensively evaluate threat degrees of different power fault types, improve the efficiency of power fault processing and ensure the safe, stable and reliable operation of a power system.
Description
Technical Field
The invention relates to the technical field of power grid fault assessment, in particular to a power grid fault threat degree assessment method, a device, equipment and a storage medium.
Background
The power system is an essential infrastructure necessary for modern society, and its reliable operation is critical for maintaining normal operation of the society. However, power failure is an unavoidable problem in power systems, and the occurrence of failure may cause damage to equipment, interruption of power supply, and even safety accidents. Therefore, in the power system, detection and evaluation of power faults are very important, wherein extraction of power fault characteristics and threat level evaluation are critical.
The fault event in the power system may be accompanied by the generation of a large amount of data, the power fault features are complex and various, multiple parameters are involved, the power system is generally affected by multiple factors such as the running state of the power grid, the load change and the like, and different types of power fault manifestations are different, so that the complexity of the power fault features is further increased. The prior art cannot accurately extract key power fault characteristics from mass data in real time, and is difficult to accurately evaluate threat degrees of different fault types, so that the processing efficiency of power faults is low, and the stability, safety and reliability of a power system are affected.
Disclosure of Invention
The invention aims to provide a power grid fault threat degree assessment method, device, equipment and storage medium, which are used for solving the technical problem that the threat degrees of different power fault types are difficult to accurately assess to reduce the fault processing efficiency in the prior art.
The aim of the invention can be achieved by the following technical scheme:
in a first aspect, a method for evaluating a threat level of a power grid fault includes the steps of:
Acquiring power utilization time sequence data of a power grid user in a preset period, and converting the power utilization time sequence data into corresponding electric image data;
Acquiring space-time feature vectors corresponding to the electricity consumption image data according to a preset feature mining model, and identifying a plurality of electricity consumption fault types to be evaluated according to the space-time feature vectors;
Determining the relative priority corresponding to each power consumption fault type according to a preset fault threat degree evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
And evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority.
Optionally, the feature mining model includes a feature extraction network and a gated recurrent neural network, and the acquiring the space-time feature vector in the electrical image data according to a preset feature mining model includes:
extracting spatial features in the electrical image data according to the feature extraction network to obtain corresponding spatial feature vectors;
and extracting time characteristics in the electrical image data according to the gating cyclic neural network to obtain corresponding time characteristic vectors.
Optionally, the feature extraction network includes a convolution layer, a pooling layer and a full connection layer, and the extracting, according to the feature extraction network, spatial features in the electrical image data to obtain corresponding spatial feature vectors includes:
Performing convolution operation on the electrical image data by using the convolution layer to obtain a corresponding spatial feature map;
Carrying out pooling operation on the spatial feature map by utilizing the pooling layer to obtain the simplified spatial feature map;
and integrating the simplified spatial feature patterns by using the full connection layer to obtain corresponding spatial feature vectors.
Optionally, the identifying a plurality of types of electrical faults to be evaluated according to the space-time feature vector includes:
And inputting the space-time feature vector into a regressor or a classifier, predicting the electricity consumption behavior of the user or detecting the abnormal electricity consumption behavior of the user, and identifying and obtaining a plurality of electricity consumption fault types to be evaluated.
Optionally, the determining the relative priority corresponding to each power consumption fault type according to a preset fault threat degree evaluation model includes:
constructing a fault threat assessment index set according to a preset fault threat degree assessment model;
Determining the index weight corresponding to each fault threat assessment index in the fault threat assessment index set based on an objective weighting algorithm;
and determining the relative priority corresponding to each power utilization fault type according to the fault threat assessment index set and each index weight.
Optionally, the determining, based on the objective weighting algorithm, an index weight corresponding to each fault threat assessment index in the fault threat assessment index set includes:
calculating correlation coefficients among all fault threat assessment indexes in the fault threat assessment index set;
respectively calculating standard deviation and information weight of each fault threat assessment index;
and calculating the index weight of each fault threat assessment index according to the standard deviation and the information weight.
Optionally, the determining, according to the fault threat assessment index set and each index weight, a relative priority corresponding to each power consumption fault type includes:
acquiring initial decision matrixes of the fault threat assessment indexes at different moments, calculating time sequence weights corresponding to the fault threat assessment indexes by adopting an inverse poisson distribution method, and constructing a dynamic decision matrix according to the initial decision matrixes and the time sequence weights;
Constructing a mean value matrix and a standard deviation matrix in the dynamic decision matrix;
determining a first attribute weight of the mean matrix and a second attribute weight of the standard deviation matrix based on an objective weighting algorithm;
According to the first attribute weight and the second attribute weight, respectively calculating a first relative weight of the mean matrix and a second relative weight of the standard deviation matrix;
and determining the relative priority corresponding to each power utilization fault type according to the first relative weight and the second relative weight.
In a second aspect, a power grid fault threat level assessment apparatus includes:
The data conversion module is used for acquiring power utilization time sequence data of a power grid user in a preset period and converting the power utilization time sequence data into corresponding power utilization image data;
The fault type identification module is used for acquiring space-time feature vectors corresponding to the electricity consumption image data according to a preset feature mining model and identifying a plurality of electricity consumption fault types to be evaluated according to the space-time feature vectors;
the fault priority determining module is used for determining the relative priority corresponding to each power utilization fault type according to a preset fault threat degree evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
And the fault evaluation module is used for evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority.
In a third aspect, a computer device comprises: a processor and a memory;
The memory stores a computer program, and the processor realizes the steps of the power grid fault threat degree evaluation method when executing the computer program.
In a fourth aspect, a computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the grid fault threat level assessment method.
Based on the technical scheme, the invention has the beneficial effects that:
Firstly, acquiring one-dimensional electricity utilization time sequence data of a power grid user in a preset period, and converting the one-dimensional electricity utilization time sequence data into two-dimensional electricity utilization image data; then, the spatial features and the temporal features among different electric image data are mined by utilizing the feature mining model, so that the spatial relationship among the different electric image data can be more comprehensively understood, and the system is improved in grasping the overall electric power use mode; the feature mining model accurately extracts the time features of the high-dimensional nonlinear time sequence, and is helpful for better understanding and analyzing the evolution process of the user electricity behavior; according to the high-dimensional space-time feature vector extracted by the feature mining model, the prediction of the electricity utilization behavior of a specific user and the detection of the abnormal electricity utilization behavior of the user are realized, and the method has high practical application value. The fault threat index attribute set is constructed according to the fault threat degree evaluation model, so that various uncertainty factors are considered more comprehensively, and the robustness of the fault threat degree evaluation model is improved; the influence of the overall change trend of the fault information on the fault threat degree evaluation can be accurately and comprehensively measured, so that the power system can more flexibly cope with different dynamic environments. The relative priority of each fault threat degree evaluation index determined by the fault threat degree evaluation model can objectively reflect the importance of different threat degree evaluation indexes, so that the accuracy of threat degree evaluation results is improved, the fault threat degree results are more reliable, the priority processing sequence of the faults is determined according to the fault threat degree evaluation results, and the fault processing efficiency can be improved.
The embodiment of the invention can efficiently and accurately extract key power fault characteristics, comprehensively and comprehensively evaluate the threat degrees of different power fault types, reasonably determine the priority order of fault processing according to the threat degrees, improve the efficiency of power fault processing and ensure the safe, stable and reliable operation of a power system.
Drawings
FIG. 1 is a schematic flow chart of a first embodiment of a method for evaluating the threat level of a power grid fault;
FIG. 2 is a schematic flow chart of a second embodiment of a method for evaluating the threat level of a power grid fault;
FIG. 3 is a schematic structural diagram of a feature extraction network in a first embodiment of a method for evaluating the threat level of a power grid fault according to the present invention;
FIG. 4 is a schematic structural diagram of a feature extraction model in a first embodiment of a power grid fault threat level assessment method according to the present invention;
FIG. 5 is a schematic diagram of a time step in an embodiment of a method for evaluating a threat level of a power grid according to the present invention;
FIG. 6 is a schematic flow chart of a fault threat level assessment model in a first embodiment of a power grid fault threat level assessment method of the present invention;
Fig. 7 is a schematic structural diagram of an embodiment of the power grid fault threat level assessment device of the present invention.
Detailed Description
The embodiment of the invention provides a power grid fault threat degree assessment method, device, equipment and storage medium, which are used for solving the technical problem that the threat degrees of different power fault types are difficult to accurately assess to reduce the fault processing efficiency in the prior art.
In order that the invention may be readily understood, a more complete description of the invention will be rendered by reference to the appended drawings. Preferred embodiments of the present invention are shown in the drawings. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
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 herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.
The power failure characteristics are complex and various, and a plurality of parameters such as voltage, current, frequency and the like are involved, and the dynamic change of the parameters in the power grid makes accurate extraction and quantification of the failure characteristics difficult. Different types of faults may exhibit different waveform and spectral characteristics, which are often affected by various factors such as grid operating conditions, load variations, etc., increasing the complexity of the power fault characteristics. Fault events in a power system may be accompanied by the generation of large amounts of data, such as waveform data, event records, etc. In the prior art, key fault characteristics cannot be extracted from the mass data efficiently and accurately, and the preferred sequence of the fault processing cannot be ordered accurately, so that the efficiency of the fault processing is low.
Referring to fig. 1, the present invention provides a first embodiment of a power grid fault threat level assessment method, which includes the following steps:
s100: acquiring power utilization time sequence data of a power grid user in a preset period, and converting the power utilization time sequence data into corresponding electric image data;
S200: acquiring space-time feature vectors corresponding to the electricity consumption image data according to a preset feature mining model, and identifying a plurality of electricity consumption fault types to be evaluated according to the space-time feature vectors;
S300: determining the relative priority corresponding to each power consumption fault type according to a preset fault threat degree evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
S400: and evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority.
Firstly, power utilization time sequence data of a power grid user in a preset period are obtained, and one-dimensional power utilization time sequence data are converted into corresponding two-dimensional power utilization image data, so that spatial characteristics among different power utilization time sequence data can be captured better. When the data is converted, the user behavior input data is required to be converted into two-dimensional data from a one-dimensional user electricity load sequence in a data folding mode. Specifically, the user load data arranged in the unit of day may be converted into the unit of week. Taking the load data of a certain user for 28 days as an example, the structure of the data is 1 multiplied by 28, and the structure of the data is changed into 1 multiplied by 4 multiplied by 7 by converting the data into the data dimension by taking the week as a unit based on the power consumption behavior characteristic mining model bit observation of ConvGRU.
The two-dimensional electrical image data is used as the input data of the feature mining model, and the two reasons are as follows: first, since the daily power consumption fluctuates in a relatively independent manner, it is difficult to identify the periodic or aperiodic nature of the power usage from the one-dimensional power consumption data; secondly, by aligning the power consumption data in a preset time period (for example, 4 weeks) and inputting the power consumption data in a two-dimensional image data form, the characteristic of global perception of the feature extraction network CNN can be exerted, and the spatial characteristics among different users and among different time periods of the same user can be conveniently mined.
In one embodiment, the feature mining model may include a feature extraction network and a gated recurrent neural network, and acquiring space-time feature vectors in the electricity image data according to a preset feature mining model includes:
extracting spatial features in the power utilization image data according to the feature extraction network to obtain corresponding spatial feature vectors;
And extracting time characteristics in the power utilization image data according to the gating cyclic neural network to obtain corresponding time characteristic vectors.
The feature mining model (ConvGRU) in the embodiment of the invention is a user electricity behavior feature mining model, can perform feature mining on the electricity behavior of a user, extracts spatial features among different user data by utilizing the feature extraction network (CNN) capable of globally sensing, and mines time features inside the electricity data by utilizing a gate-controlled cyclic neural network (GRU). The feature mining model comprehensively considers the spatial global features and the temporal local features of the user electricity data, and can improve the reliability and the representativeness of feature mining. Meanwhile, in order to improve the learning efficiency of the electricity consumption behavior feature mining model and enable parameters and training of the model to be more targeted, the embodiment of the invention can also introduce a clustering algorithm such as a K-means clustering method to perform clustering analysis on the electricity consumption data of the user, and then train the corresponding feature mining model according to different categories.
The user electricity behavior feature mining model in the embodiment of the invention can comprehensively consider the spatial features and the temporal features of the user electricity data so as to improve the reliability and the representativeness of feature mining. The feature mining model mainly comprises three steps:
step 1: spatial feature extraction
After the one-dimensional user power consumption time sequence data is converted into the two-dimensional power consumption image data, the spatial characteristics among different power consumption time sequence data can be better captured. The spatial features in the two-dimensional electrical image data are extracted by a feature extraction network CNN (also called an encoder) to form a spatial feature map containing the spatial data features.
Step 2: temporal feature mining
And (3) the spatial feature map extracted by the encoder is expanded through a time scale to form an intermediate vector which accords with the input form of the gating cyclic neural network, and then the intermediate vector is transmitted to a decoder for time feature mining. And (3) mining the time characteristics of the intermediate characteristic vector extracted by the encoder by using a gated cyclic neural network (GRU neural network) (also called a decoder). The GRU is trained using time back propagation by resetting gates and updating the time dependencies between the gate learning data.
Step 3: task execution and application
And obtaining space-time feature vectors corresponding to the electricity utilization image data according to the extracted time features and the spatial features, wherein the feature mining model can execute different tasks and can comprise the prediction of electricity utilization behaviors and the detection of abnormal electricity utilization behaviors. For predictive problems, linear regression may be employed; for the problem of anomaly detection, logistic regression is introduced as a classifier, and the network and the logistic regression are jointly involved in training. FIG. 4 illustrates a block diagram of a user electrical behavior feature mining model based on ConvGRU.
The feature mining model in the embodiment of the invention integrates the advantages of the feature extraction network CNN and the gating cyclic neural network GRU, and is suitable for diversified power utilization behavior analysis tasks by extracting time features and space features of the electrical image data for mining.
In one embodiment, the feature extraction network CNN may include a convolution layer, a pooling layer, and a full connection layer, and the feature extraction network extracts spatial features in the electricity image data according to the features to obtain corresponding spatial feature vectors, including:
Carrying out convolution operation on the electrical image data by utilizing the convolution layer to obtain a corresponding space feature map;
carrying out pooling operation on the space feature map by utilizing the pooling layer to obtain a simplified space feature map;
And integrating the simplified spatial feature patterns by using the full connection layer to obtain corresponding spatial feature vectors.
The feature extraction network in embodiments of the present invention is a neural network suitable for processing computer vision images, and may be comprised of a convolutional layer, a sampling layer, and a fully-connected layer. The feature extraction network can learn effective spatial features from original input images, namely the electrical image data by using a deep framework, and the specific structure of the feature extraction network is shown in fig. 3.
Specifically, the convolution layer is a core structure of the feature extraction network, and may include two convolution layers, where each convolution layer includes a certain number of feature traps (also called feature extractors), and when input electrical image data is transferred to the convolution layer, each feature trap performs convolution operation along sliding and translation of the image to generate a spatial feature map corresponding to the image; with the stacking of the convolution layers, the deeper spatial features implicit in the input will be extracted and learned; the spatial feature map is obtained after the input matrix is subjected to the translational convolution of the feature catcher. Fig. 3 shows a basic structure of a two-dimensional convolution network, for example, a 3×3 feature grabber, where the input matrix is subjected to a translational convolution of the feature grabber to obtain a feature map. The ith input to layer 1-1The convolution output of the next layer is shown in the formula (1):
Where f (·) is represented as an activation function for the convolutional layer, N M is all selected inputs; Representing a convolution operation,/> Is a paranoid item; /(I)The coefficient matrix between the i-th input and the j-th input is connected for the i-th layer.
It should be noted that each coefficient matrix can learn only a single feature, so multiple feature collectors are required in one convolution layer, and deep features in the data need to be captured by stacking more convolution layers. One of the limitations of conventional neural networks is the poor scalability due to the complete connection of neurons. The feature extraction network in embodiments of the present invention overcomes the shortcomings of conventional neural networks by connecting each neuron to its neighbors (not all neurons). A local region consisting of a set of neurons, also called the perceived field of view, will be used to convolve the input 2D data with feature traps having the same perceived field of view. Thus, the convolutional layer has the ability to globally perceive the input data, rather than being limited to the size of the acceptance field.
In particular, the pooling layer, also referred to as a sub-sampling layer, is typically used to reduce the training parameters (e.g., training weights and number of feature extractors) of the feature extraction network and the number of reduced feature maps immediately following the convolution layer. In addition, the pooling layer can also be used to control the convergence of the neural network to avoid overfitting during training. Maximum pooling is one of the most typical pooling operations, i.e. maximizing feature points within a neighborhood. Model parameters after the pooling layer are reduced, and the calculation speed is improved. The operation of the pooling layer can be expressed as formula (2):
Wherein, The multiplicative deviation coefficient representing the jth output map of the layer l, down (·) is represented as a sub-sampling process in pooling, with different expressions depending on the pooling technique.
Taking the maximum value pool of 2×2 sizes as shown in fig. 3 as an example, down (·) represents that the feature map of the previous layer is taken as input, and the maximum value thereof is selected for each region of 2×2 sizes as output. Essentially, there are two directional delivery processes by the pooling layer in the feature extraction network: 1) During forward pass, the pooling layer will select the maximum of the area covered by the pooling filter; 2) During the back propagation, the pooling layer will send the maximum of the gradient from back to front to the front feature grabber (also called feature extractor).
Specifically, the fully connected layer is used for integrating the spatial feature patterns extracted by the plurality of feature extractors and reduced by the maximum pooling layer so as to enable the spatial feature patterns to exist in the form of spatial feature vectors. The fully connected layer maps the learned spatial feature representation to the labeled space of the sample, thereby providing a basis for further classification or regression problems.
In the embodiment of the invention, the feature extraction part of the feature mining model adopts the structure of an encoder and a decoder, the encoder is a feature extraction network for carrying out space feature extraction and dimension reduction on the input two-dimensional electrical image data, and the decoder is a gate control cyclic neural network GRU for carrying out time feature extraction on the processed intermediate feature vector. After the original input electrical image data are transmitted into a CNN feature extraction network, a spatial feature map containing spatial data features is formed through the processing of a feature extractor and a pooling layer; then, the spatial feature pattern acquired by the encoder is expanded through a time scale to form an intermediate vector which accords with the input form of the decoder, the intermediate vector is used as input data to be transmitted to the gate control cyclic neural network GRU, and the reset gate and the update gate in the gate control cyclic neural network GRU continuously adjust own parameters in a large amount of training, so that the time dependence between the intermediate vector and the electrical image data can be learned.
It will be appreciated that the feature extraction network CNN is an encoder and the gated recurrent neural network GRU is a decoder.
Specifically, referring to fig. 4 and 5, the encoder includes 2 convolutional layers, each of which may have 64 feature extractors therein, the feature extractor of the first convolutional layer has a size of 4×4, the feature extractor of the second convolutional layer has a size of3×3, and the feature extractors share parameters throughout the time-frequency space. The pooling strategy in the encoder uses a non-overlapping maximum pooling method of size 3 x 3 and is only used after the first convolutional layer. The spatial feature map extracted by the encoder is expanded through a time scale to form an input form conforming to a gating cyclic neural network, and then the intermediate vector is transferred to a GRU in a decoder for time feature mining.
Specifically, the decoder includes 2 GRU layers, each of which may include 100 nerve units, and each GRU layer is followed by a Dropout layer, and the Dropout layer is used for thinning to enhance the generalization capability of the model. The GRU of the decoding layer is trained by adopting a time back propagation method, and a time step expansion schematic diagram of the whole ConvGRU model is shown in fig. 5. In fig. 5, the procedure from the input data to the intermediate feature s= (S 0,…,ST) can be found by convolution and pooling. The space-time feature vector p= (P 0,…,PT) is obtained after passing through the GAU layer from the intermediate feature vector S, so that on the basis of the user electricity behavior characteristics, the prediction (regression problem) of the user electricity behavior and the identification (classification problem) of the user abnormal electricity behavior can be further completed.
Specifically, for the problem of predicting the electricity consumption behavior of the user, the method can be completed by linear regression on the basis of the characteristics of the electricity consumption behavior. For abnormal electricity behavior detection, the problem can be converted into a classification problem, and the classification problem is generally obtained by accessing logistic regression (Logistic Regression, LR) at the end of the network as a classifier of a model to participate in training together with the network. The training set for the set model is { (x 1,y1),(x2,y2),…,(xn,yn) }, where y n ε {0,1}, n is the number of samples. As a mapping function of logistic regression, the Sigmoid function can well map the classification result to between [0,1 ]. The Sigmoid function is shown in formula (3):
Where θ is the parameter that the model needs to train. The Sigmoid mapping function represents the probability P that the result belongs to category y=1. Thus, the probabilities of the classification classes 0 and 1 can be shown as formula (4) and formula (5), respectively:
P(y＝0∣x；θ)＝1-gθ(x)；(4)
P(y＝1∣x；θ)＝gθ(x)；(5)
From the maximum likelihood estimation, the loss function of the available model is the cross-over, as shown in equation (6):
by utilizing the feature mining model in the embodiment of the invention, the time features and the space features (time-space features for short) corresponding to the user electricity behavior can be fully extracted, and the prediction and the identification of the electricity failure types in the user electricity behavior are completed according to the time-space features of the user electricity behavior, so that a plurality of electricity failure types to be evaluated are obtained.
In order to further judge the threat degree of the power consumption fault type so as to determine the priority processing sequence of the power consumption fault type, the embodiment of the invention determines the relative priority processing sequence among different power consumption fault types by constructing a fault threat degree evaluation model so as to improve the efficiency of power consumption fault processing.
In one embodiment, determining the relative priority corresponding to each electrical fault type according to a preset fault threat level assessment model includes:
constructing a fault threat assessment index set according to a preset fault threat degree assessment model;
Determining index weights corresponding to all fault threat assessment indexes in the fault threat assessment index set based on an objective weighting algorithm;
and determining the relative priority corresponding to each power failure type according to the failure threat assessment index set and each index weight.
In one embodiment, determining an index weight corresponding to each of the failure threat assessment indexes in the failure threat assessment index set based on the objective weighting algorithm includes:
Calculating correlation coefficients among all fault threat assessment indexes in the fault threat assessment index set;
Respectively calculating standard deviation and information weight of each fault threat assessment index;
And calculating the index weight of each fault threat assessment index according to the standard deviation and the information weight.
In the embodiment of the invention, firstly, the fault threat assessment index attribute set is constructed by considering the fault detection information characteristics influencing the fault threat degree assessment. On the basis, the probability of detecting information in the fault threat degree evaluation process is characterized by the triangular probability distribution.
Secondly, constructing a probability distribution mean value matrix and a probability standard deviation matrix so as to measure the influence of the overall change trend of fault information on the evaluation of the degree of fault threat, wherein the greater the fluctuation of fault attribute information is, the greater the risk of threat caused by the fault is;
In addition, a fault threat degree evaluation index weight determining model based on an objective weighting algorithm is provided, and the influence of conflicts among fault threat degree evaluation indexes on the index weight is reflected;
Finally, a model for expanding multi-criterion decision based on the probability distribution is used for obtaining the joint dominance of the faults through calculation, so that a fault threat result can be reflected, the dynamic requirement of fault threat degree assessment is realized, and the reliability of an auxiliary decision result is improved.
In order to consider the influence of the internal variation degree among fault threat degree evaluation indexes on the fault type attribute weight and the influence of the conflict among fault type attributes on the threat evaluation result, the embodiment of the invention utilizes a fault type attribute weight algorithm based on an objective weighting algorithm to determine the attribute weight (index weight) of the fault threat degree evaluation.
Specifically, assuming that m power consumption fault types to be evaluated exist, and n fault threat evaluation indexes form a decision matrix a= (a) m×n; the method for determining the index weight of each fault threat assessment index pair based on the objective weighting algorithm mainly comprises the following steps:
Step 1: calculating a correlation coefficient rho jk between fault threat assessment indexes of the power consumption fault types, wherein the correlation coefficient rho jk is shown as a formula (7):
Wherein, Ρ jk represents a correlation coefficient between jth and kth fault threat assessment indexes, a ij represents a jth fault threat assessment index value of an ith power consumption fault type, a ik represents a kth fault threat assessment index value of an ith power consumption fault type,/>Represents the average value of the j-th fault threat assessment indicator,Represents the average of the kth fault threat assessment indicators.
Step 2: calculating standard deviation sigma j and information bearing capacity (information weight) C j of all fault threat assessment indexes, as shown in the formulas (8) and (9);
Wherein: sigma j is the standard deviation of the j-th fault threat level evaluation index attribute, and can be used as the measurement of the comparison intensity of the fault threat level evaluation index attribute; ρ jk is a correlation coefficient between the fault threat level assessment index attribute j and the fault threat level assessment index attribute k; while The conflict among the attribute indexes of the fault threat degree evaluation index is represented;
step 3: when the fault threat degree evaluation index attribute C j is larger, the information which influences the threat evaluation result and is contained in the j-th fault threat degree evaluation index attribute is more, and the corresponding weight is higher; calculating attribute weight w j of the j-th fault threat degree evaluation index attribute of each power consumption fault type according to the formula (10);
In one embodiment, determining the relative priority corresponding to each electrical fault type based on the set of fault threat assessment indicators and each indicator weight includes:
Acquiring initial decision matrixes of all fault threat assessment indexes at different moments, calculating time sequence weights corresponding to all the fault threat assessment indexes by adopting an inverse poisson distribution method, and constructing a dynamic decision matrix according to the initial decision matrixes and the time sequence weights;
constructing a mean value matrix and a standard deviation matrix in the dynamic decision matrix;
determining a first attribute weight of the mean matrix and a second attribute weight of the standard deviation matrix based on an objective weighting algorithm;
According to the first attribute weight and the second attribute weight, respectively calculating a first relative weight of the mean matrix and a second relative weight of the standard deviation matrix;
and determining the relative priority corresponding to each power failure type according to the first relative weight and the second relative weight.
The embodiment of the invention combines the advantages of multi-criterion decision and triangular possibility distribution, expands a multi-criterion decision algorithm based on the possibility distribution, and builds a fault threat degree evaluation model based on the possibility distribution expanded multi-criterion decision, and a flow chart of the model is shown in figure 6. The fault threat degree evaluation model determines the relative priority corresponding to each power consumption fault type, and the specific steps are as follows:
step 1: and acquiring initial decision matrixes of fault threat degree evaluation indexes at different moments.
The initial decision matrix at the moment when the set of m fault types to be evaluated is set to be x= { X i |i=1, 2, …, m }, the set of fault threat assessment index attributes is c= { C j∣j＝1,2,…,n},tk is denoted by a k, wherein the assessment value of the power consumption fault type X m under the assessment index attribute C n at the moment t k is a triangle likelihood distribution, that is (a mn-αmn,amn,amn+βmn).Ak is shown as formula (11):
Step 2: and calculating the time sequence weight corresponding to each fault threat assessment index by adopting an inverse poisson distribution method.
In practical application, the closer to the decision information at the current moment, the greater the influence on the fault threat degree evaluation result. The time series weights are calculated using the inverse poisson distribution method as shown in equation (12).
Wherein: eta k is more than or equal to 0 andK represents a kth fault threat level evaluation index attribute (fault threat evaluation index for short).
Step 3: and fusing the evaluation information at different moments to obtain a dynamic decision matrix R, namely constructing the dynamic decision matrix according to the initial decision matrix and the time sequence weight.
The decision information after fusing the plurality of time evaluation information is expressed as: r= (R m1,rm2,…,rmn) as shown in formula (13):
step 4: and constructing a probability mean value matrix and a probability standard deviation matrix in the dynamic decision matrix R.
Since each element in R is characterized by a triangular likelihood distribution, each element can be calculated as a mean M (R ij) and a standard deviation StD (R ij), as shown in equations (14) and (15):
The probability mean matrix and the probability standard deviation matrix obtained by the above-described formula construction are shown in the formula (16) and the formula (17):
Step 5: the attribute weight is the key of fault threat degree assessment, and the first attribute weight W M of the probability mean matrix and the second attribute weight theta SID of the probability standard deviation matrix are calculated by using a fault type attribute determination algorithm based on objective weighting, as shown in the formula (18) and the formula (19):
WM＝[w1 w2…wn]； (18)
θsID＝[θ1 θ2…θn]； (19)
g is [1, n ], u and k are [1, m ].
Step 6: respectively calculating first relative weights of a probability mean matrix and a decision information probability standard deviation matrix of the dynamic decision matrix RAnd a second relative weight/>
When M (r) ug>M(r)kg, the transition probability weight is as shown in formula (20):
m (r) ug and M (r) kg are threat indicators for fault types u and k, respectively; delta is a constant.
And the attribute weight value is the attribute weight value of the probability mean matrix.
Conversely, when M (r) ug<M(r)kg, the transition probability weight is as shown in formula (21):
When the probability standard deviation Std (r) ug>Std(r)kg, the transition probability weight is as shown in formula (22):
Conversely, when the likelihood standard deviation Std (r) ug<Std(r)kg, the transition probability weight is as shown in formula (23):
attribute weight values representing the likelihood standard deviation matrix.
It will be appreciated that the number of components,Includes/>And/>ComprisingAnd/>
Step 7: calculating dominance of fault type X u relative to other fault typesAnd/>As shown in the formulas (24) and (25).
Wherein,
M (I u,Ik) and StD (I u,Ik) represent a probability mean matrix and a probability standard deviation matrix of the power failure types I u and I k, respectively, and ω and ζ are constants.
Step 8: calculating the overall dominance omega M(xu of the probability average matrix and the overall dominance omega StD(xu of the probability standard deviation matrix of threat results of each power failure type relative to other power failure types) respectively, as shown in the formula (28) and the formula (29):
Step 9: the joint dominance of each fault type over the other fault types is calculated as shown in equation (30).
Ω(xu)＝((ΩM(xu))∧α+(ΩStD(xu)∧β)/2；(30)
Where α+β=1, α represents a central change trend of the power consumption failure type information (a duty ratio of a probability mean matrix evaluation result of the power consumption failure type), and β represents an importance degree of the magnitude of fluctuation (a duty ratio of a probability standard deviation matrix evaluation result of the power consumption failure type). When α=1, β=0, the threat level of the power failure type is determined only by the centralized change trend of the power failure type information; and α=0, β=1, then the threat levels of the respective power failure types are ranked by using the power failure type fluctuation information. The values of alpha and beta can be dynamically adjusted according to the judgment of the situation, and the corresponding threat ordering result of the power failure type can be obtained.
Firstly, acquiring one-dimensional electricity utilization time sequence data of a power grid user in a preset period, and converting the one-dimensional electricity utilization time sequence data into two-dimensional electricity utilization image data; then, the spatial features and the temporal features among different electric image data are mined by utilizing the feature mining model, so that the spatial relationship among the different electric image data can be more comprehensively understood, and the system is improved in grasping the overall electric power use mode; the feature mining model accurately extracts the time features of the high-dimensional nonlinear time sequence, and is helpful for better understanding and analyzing the evolution process of the user electricity behavior; according to the high-dimensional space-time feature vector extracted by the feature mining model, the prediction of the electricity utilization behavior of a specific user and the detection of the abnormal electricity utilization behavior of the user are realized, and the method has high practical application value. The fault threat index attribute set is constructed according to the fault threat degree evaluation model, so that various uncertainty factors are considered more comprehensively, and the robustness of the fault threat degree evaluation model is improved; the influence of the overall change trend of the fault information on the fault threat degree evaluation can be accurately and comprehensively measured, so that the power system can more flexibly cope with different dynamic environments. The relative priority of each fault threat degree evaluation index determined by the fault threat degree evaluation model can objectively reflect the importance of different threat degree evaluation indexes, so that the accuracy of threat degree evaluation results is improved, the fault threat degree results are more reliable, the priority processing sequence of the faults is determined according to the fault threat degree evaluation results, and the fault processing efficiency can be improved.
The embodiment of the invention can efficiently and accurately extract key power fault characteristics, comprehensively and comprehensively evaluate the threat degrees of different power fault types, reasonably determine the priority order of fault processing according to the threat degrees, improve the efficiency of power fault processing and ensure the safe, stable and reliable operation of a power system.
Referring to fig. 2, the present invention provides a second embodiment of a power grid fault threat level assessment method, which includes the following steps:
Step S1: constructing a feature mining model, converting 1-dimensional user electricity data into 2-dimensional data by the feature mining model, and mining the spatial characteristics among different electricity sequences by utilizing the characteristics of global perception of a feature extraction network in the feature mining model; the characteristic indexes extracted by the characteristic extraction network are further processed through a gating circulating neural network in a characteristic mining model, and the advantages of the gating circulating neural network in processing the high-dimensional nonlinear time sequence are utilized to extract time characteristics; the extracted high-dimensional feature vectors in the feature mining model are integrated through a full connection layer and sent into a regressor or a classifier to complete specific electricity consumption behavior prediction and identification tasks;
Step S2: constructing a fault threat degree evaluation model, and constructing a fault threat index attribute set based on fault detection information characteristics influencing the fault threat degree evaluation in the fault threat degree evaluation model; and the probability of detecting information in the fault threat degree evaluation process is characterized by using the triangular probability distribution; constructing a probability distribution mean matrix and a probability standard deviation matrix so as to measure the influence of the overall change trend of fault information on the fault threat degree evaluation, and determining a model based on the fault threat degree evaluation index weight of the objective weighting algorithm to reflect the influence of conflict among fault threat degree evaluation indexes on the index weight; and (3) expanding a model of multi-criterion decision based on the probability distribution, obtaining the joint dominance of the faults through calculation, and obtaining a result reflecting the threat degree of the faults.
The power grid fault threat degree assessment method provided by the embodiment of the invention comprises the following steps: constructing a feature extraction model, integrating high-dimensional feature vectors extracted by the feature extraction model through a full-connection layer, and sending the high-dimensional feature vectors into a regressor or a classifier to complete specific electricity behavior prediction and identification tasks; constructing a fault threat degree evaluation model, and constructing a fault threat index attribute set based on fault detection information characteristics influencing the fault threat degree evaluation in the fault threat degree evaluation model; and (3) expanding a model of multi-criterion decision based on the probability distribution, obtaining the joint dominance of the faults through calculation, and obtaining a result reflecting the threat degree of the faults. Threat degrees of different power fault types can be comprehensively and comprehensively evaluated, so that a fault threat degree result is more reliable, and the fault processing efficiency can be improved.
According to the embodiment of the invention, one-dimensional user electricity utilization time sequence data are converted into two-dimensional electricity utilization image data, the spatial characteristics among different user electricity utilization sequence data are mined by utilizing the characteristic extraction network in the characteristic mining model, and the overall perception characteristic of the characteristic extraction network is utilized, so that the relation among different electricity utilization sequence data can be understood more comprehensively, and the system can grasp the overall electricity utilization mode; the time characteristics of the high-dimensional nonlinear time sequence are accurately extracted by using the gating circulating neural network in the characteristic mining model, so that the evolution process of the electricity consumption behavior can be better understood and analyzed; the high-dimensional feature vectors extracted by the feature mining model are integrated through the full connection layer and then sent into a regressor or a classifier, so that the prediction and identification tasks of specific electricity utilization behaviors are realized, and the model has higher practicability in practical application. Constructing a fault threat index attribute set based on fault detection information characteristics influencing fault threat degree evaluation, and providing effective input data for subsequent evaluation; the probability of detection information in the fault threat degree evaluation process is represented by using the triangular probability distribution, so that uncertainty factors are considered more comprehensively, and the robustness of the model is improved; and constructing a probability distribution mean value matrix and a probability standard deviation matrix so as to measure the influence of the overall change trend of fault information on the fault threat degree evaluation, so that the system can more flexibly cope with different dynamic environments. The index weight of each fault threat degree evaluation index is determined based on the objective weighting algorithm, so that the importance of different indexes can be objectively reflected, and the accuracy of the evaluation result is improved. Based on a multi-criterion decision model with expanded probability distribution, the joint dominance (namely relative priority) of the faults is obtained through calculation, more comprehensive guidance is provided for comprehensive evaluation, the fault threat degree result is more reliable, and the fault processing efficiency is improved. The embodiment of the invention can efficiently and accurately extract key power fault characteristics, comprehensively and comprehensively evaluate the threat degrees of different power fault types, reasonably determine the priority order of fault processing according to the threat degrees, improve the efficiency of power fault processing and ensure the safe, stable and reliable operation of a power system.
Referring to fig. 7, the present invention provides an embodiment of a power grid fault threat level assessment apparatus, including:
the data conversion module 11 is used for acquiring power utilization time sequence data of a power grid user in a preset period and converting the power utilization time sequence data into corresponding power utilization image data;
The fault type recognition module 22 is configured to obtain a space-time feature vector corresponding to the electrical image data according to a preset feature mining model, and recognize a plurality of electrical fault types to be evaluated according to the space-time feature vector;
The fault priority determining module 33 is configured to determine a relative priority corresponding to each of the power consumption fault types according to a preset fault threat level evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
The fault evaluation module 44 is configured to evaluate the degree of fault threat of the corresponding power consumption fault type according to each of the relative priorities.
Firstly, acquiring one-dimensional electricity utilization time sequence data of a power grid user in a preset period, and converting the one-dimensional electricity utilization time sequence data into two-dimensional electricity utilization image data; then, the spatial features and the temporal features among different electric image data are mined by utilizing the feature mining model, so that the spatial relationship among the different electric image data can be more comprehensively understood, and the system is improved in grasping the overall electric power use mode; the feature mining model accurately extracts the time features of the high-dimensional nonlinear time sequence, and is helpful for better understanding and analyzing the evolution process of the user electricity behavior; according to the high-dimensional space-time feature vector extracted by the feature mining model, the prediction of the electricity utilization behavior of a specific user and the detection of the abnormal electricity utilization behavior of the user are realized, and the method has high practical application value. The fault threat index attribute set is constructed according to the fault threat degree evaluation model, so that various uncertainty factors are considered more comprehensively, and the robustness of the fault threat degree evaluation model is improved; the influence of the overall change trend of the fault information on the fault threat degree evaluation can be accurately and comprehensively measured, so that the power system can more flexibly cope with different dynamic environments. The relative priority of each fault threat degree evaluation index determined by the fault threat degree evaluation model can objectively reflect the importance of different threat degree evaluation indexes, so that the accuracy of threat degree evaluation results is improved, the fault threat degree results are more reliable, the priority processing sequence of the faults is determined according to the fault threat degree evaluation results, and the fault processing efficiency can be improved.
The embodiment of the invention can efficiently and accurately extract key power fault characteristics, comprehensively and comprehensively evaluate the threat degrees of different power fault types, reasonably determine the priority order of fault processing according to the threat degrees, improve the efficiency of power fault processing and ensure the safe, stable and reliable operation of a power system.
In addition, the invention also provides a computer device, which comprises: a processor and a memory;
The memory stores a computer program, and the processor realizes the steps of the power grid fault threat degree evaluation method when executing the computer program.
Meanwhile, the invention also provides a computer readable storage medium, on which a computer program is stored, and the computer program realizes the steps of the power grid fault threat degree evaluation method when being executed by a processor.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
In the embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of elements is merely a logical functional division, and there may be additional divisions of actual implementation, e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The power grid fault threat level assessment method is characterized by comprising the following steps of:
Acquiring power utilization time sequence data of a power grid user in a preset period, and converting the power utilization time sequence data into corresponding electric image data;
Acquiring space-time feature vectors corresponding to the electricity consumption image data according to a preset feature mining model, and identifying a plurality of electricity consumption fault types to be evaluated according to the space-time feature vectors;
Determining the relative priority corresponding to each power consumption fault type according to a preset fault threat degree evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
And evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority.
2. The power grid fault threat level assessment method according to claim 1, wherein the feature mining model comprises a feature extraction network and a gated recurrent neural network, the acquiring the space-time feature vector in the electrical image data according to a preset feature mining model comprises:
extracting spatial features in the electrical image data according to the feature extraction network to obtain corresponding spatial feature vectors;
and extracting time characteristics in the electrical image data according to the gating cyclic neural network to obtain corresponding time characteristic vectors.
3. The power grid fault threat level assessment method according to claim 2, wherein the feature extraction network comprises a convolution layer, a pooling layer and a full connection layer, the extracting spatial features in the electrical image data according to the feature extraction network to obtain corresponding spatial feature vectors comprises:
Performing convolution operation on the electrical image data by using the convolution layer to obtain a corresponding spatial feature map;
Carrying out pooling operation on the spatial feature map by utilizing the pooling layer to obtain the simplified spatial feature map;
and integrating the simplified spatial feature patterns by using the full connection layer to obtain corresponding spatial feature vectors.
4. The power grid fault threat level assessment method according to claim 1, wherein the identifying a number of power usage fault types to be assessed according to the space-time feature vector comprises:
And inputting the space-time feature vector into a regressor or a classifier, predicting the electricity consumption behavior of the user or detecting the abnormal electricity consumption behavior of the user, and identifying and obtaining a plurality of electricity consumption fault types to be evaluated.
5. The power grid fault threat level assessment method according to claim 1, wherein the determining the relative priority corresponding to each power consumption fault type according to a preset fault threat level assessment model includes:
constructing a fault threat assessment index set according to a preset fault threat degree assessment model;
Determining the index weight corresponding to each fault threat assessment index in the fault threat assessment index set based on an objective weighting algorithm;
and determining the relative priority corresponding to each power utilization fault type according to the fault threat assessment index set and each index weight.
6. The power grid fault threat level assessment method according to claim 5, wherein the determining, based on the objective weighting algorithm, an index weight corresponding to each fault threat assessment index in the fault threat assessment index set comprises:
calculating correlation coefficients among all fault threat assessment indexes in the fault threat assessment index set;
respectively calculating standard deviation and information weight of each fault threat assessment index;
and calculating the index weight of each fault threat assessment index according to the standard deviation and the information weight.
7. The power grid fault threat level assessment method according to claim 5, wherein the determining the relative priority corresponding to each of the power utilization fault types according to the fault threat assessment index set and each of the index weights comprises:
acquiring initial decision matrixes of the fault threat assessment indexes at different moments, calculating time sequence weights corresponding to the fault threat assessment indexes by adopting an inverse poisson distribution method, and constructing a dynamic decision matrix according to the initial decision matrixes and the time sequence weights;
Constructing a mean value matrix and a standard deviation matrix in the dynamic decision matrix;
determining a first attribute weight of the mean matrix and a second attribute weight of the standard deviation matrix based on an objective weighting algorithm;
According to the first attribute weight and the second attribute weight, respectively calculating a first relative weight of the mean matrix and a second relative weight of the standard deviation matrix;
and determining the relative priority corresponding to each power utilization fault type according to the first relative weight and the second relative weight.
8. A power grid fault threat level assessment apparatus, comprising:
The data conversion module is used for acquiring power utilization time sequence data of a power grid user in a preset period and converting the power utilization time sequence data into corresponding power utilization image data;
The fault type identification module is used for acquiring space-time feature vectors corresponding to the electricity consumption image data according to a preset feature mining model and identifying a plurality of electricity consumption fault types to be evaluated according to the space-time feature vectors;
the fault priority determining module is used for determining the relative priority corresponding to each power utilization fault type according to a preset fault threat degree evaluation model; the relative priority is used for representing the priority of the power utilization fault type relative to other power utilization fault types;
And the fault evaluation module is used for evaluating the fault threat degree of the corresponding power utilization fault type according to each relative priority.
9. A computer device, comprising: a processor and a memory;
wherein the memory stores a computer program, the processor implementing the steps of the grid fault threat level assessment method according to any one of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the grid fault threat level assessment method according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410260246.6A CN118035646A (en) | 2024-03-07 | 2024-03-07 | Power grid fault threat degree assessment method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410260246.6A CN118035646A (en) | 2024-03-07 | 2024-03-07 | Power grid fault threat degree assessment method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118035646A true CN118035646A (en) | 2024-05-14 |
Family
ID=90996806
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410260246.6A Pending CN118035646A (en) | 2024-03-07 | 2024-03-07 | Power grid fault threat degree assessment method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118035646A (en) |
-
2024
- 2024-03-07 CN CN202410260246.6A patent/CN118035646A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105224872B (en) | A kind of user's anomaly detection method based on neural network clustering | |
| CN114676742A (en) | Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network | |
| CN112905465A (en) | Machine learning model black box fairness test method and system based on deep reinforcement learning | |
| Bao et al. | Covid-gan+: Estimating human mobility responses to covid-19 through spatio-temporal generative adversarial networks with enhanced features | |
| CN114580262B (en) | Lithium ion battery health state estimation method | |
| Zu et al. | A simple gated recurrent network for detection of power quality disturbances | |
| CN115186012A (en) | Power consumption data detection method, device, equipment and storage medium | |
| Dwivedi et al. | DynamoPMU: A Physics Informed Anomaly Detection, Clustering and Prediction Method using Non-linear Dynamics on μPMU Measurements | |
| Liu et al. | How do your neighbors disclose your information: Social-aware time series imputation | |
| CN118035646A (en) | Power grid fault threat degree assessment method, device, equipment and storage medium | |
| RU135434U1 (en) | DEVICE FOR AUTOMATED SIGNAL IDENTIFICATION BASED ON NEURAL NETWORK | |
| CN111796173B (en) | Partial discharge pattern recognition method, computer device, and storage medium | |
| CN114595884A (en) | Genetic intelligent optimization neural network wind power generation equipment temperature prediction method | |
| CN114860953A (en) | Interpretable method based on few-sample relation prediction model | |
| CN113407808A (en) | Method and device for judging applicability of graph neural network model and computer equipment | |
| Prabhushankar et al. | VOICE: Variance of Induced Contrastive Explanations to quantify Uncertainty in Neural Network Interpretability | |
| CN116610783B (en) | Service optimization method based on artificial intelligent decision and digital online page system | |
| CN114611556B (en) | Multi-category motor imagery task identification method based on graph neural network | |
| CN116541794B (en) | Sensor data anomaly detection method based on self-adaptive graph annotation network | |
| CN116359762B (en) | Battery state of charge estimation method based on deep learning and network compression | |
| Shirsat et al. | Breast Cancer detection based on machine learning feature selection and extraction algorithm | |
| Wang et al. | Prediction of Battery Capacity Based on Deep Residual Network | |
| Li et al. | State-of-health Monitoring of Li-ion Battery Driven by Convolutional Neural Network and Attention Mechanism | |
| CN118395155A (en) | Deep learning-based hidden danger discharge characteristic extraction method and system for power transmission line | |
| CN118134047A (en) | Information prediction 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 |