CN117592789A - Power grid environment fire risk assessment method and equipment based on time sequence analysis - Google Patents
Power grid environment fire risk assessment method and equipment based on time sequence analysis Download PDFInfo
- Publication number
- CN117592789A CN117592789A CN202410070578.8A CN202410070578A CN117592789A CN 117592789 A CN117592789 A CN 117592789A CN 202410070578 A CN202410070578 A CN 202410070578A CN 117592789 A CN117592789 A CN 117592789A
- Authority
- CN
- China
- Prior art keywords
- data
- power grid
- time
- fire
- time sequence
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000012502 risk assessment Methods 0.000 title claims abstract description 34
- 238000012300 Sequence Analysis Methods 0.000 title claims abstract description 15
- 238000012544 monitoring process Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 12
- 238000000513 principal component analysis Methods 0.000 claims abstract description 11
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 8
- 238000010219 correlation analysis Methods 0.000 claims abstract description 8
- 238000003064 k means clustering Methods 0.000 claims abstract description 8
- 238000011156 evaluation Methods 0.000 claims abstract description 4
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000008569 process Effects 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 23
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 238000010801 machine learning Methods 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 15
- 230000000875 corresponding effect Effects 0.000 claims description 12
- 238000012731 temporal analysis Methods 0.000 claims description 11
- 238000000700 time series analysis Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 10
- 230000002596 correlated effect Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 230000007613 environmental effect Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 230000007774 longterm Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 210000004027 cell Anatomy 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 7
- 230000008859 change Effects 0.000 description 3
- 238000013480 data collection Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process 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
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Educational Administration (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of power grid environment fire risk assessment, in particular to a power grid environment fire risk assessment method and equipment based on time sequence analysis, comprising the following steps: acquiring historical fire data in a power grid environment; preprocessing data; based on the information recorded in the historical fire data time sequence, classifying fire risks by adopting an unsupervised learning K-means clustering algorithm; adopting correlation analysis and principal component analysis to select related time sequence characteristics, wherein the time sequence consists of independent variables which are power grid working parameters and environment monitoring data; constructing a time sequence data set; and inputting the test set into the model obtained through training, and carrying out model evaluation. The invention can realize comprehensive, accurate and real-time assessment of the fire risk of the power grid environment, improves the fire protection efficiency of the power grid environment, and enhances the fire protection effect of the power grid environment.
Description
Technical Field
The invention relates to the technical field of power grid environment fire risk assessment, in particular to a power grid environment fire risk assessment method and equipment based on time sequence analysis.
Background
In the power grid environment, fire risk assessment is an important and complex problem, because a fire caused by power grid faults or environmental factors may cause significant property loss and casualties, traditional risk assessment methods often depend on experience judgment and static data, and it is difficult to accurately predict and prevent fire occurrence, so it is of great significance to develop a method capable of assessing fire risk in the power grid environment in real time and dynamically, in the power grid environment, fire risk assessment faces many challenges such as data diversity, processing complexity and dynamic change, and existing methods generally lack comprehensive analysis capability on various data sources, cannot effectively process space and time characteristics of time series data, and are difficult to realize real-time updating. According to the invention, by integrating historical fire data, power grid parameters and environment monitoring data, time series data are constructed and analyzed by a machine learning method, so that comprehensive, accurate and real-time assessment of the fire risk of the power grid environment is realized.
Therefore, the application provides a power grid environment fire risk assessment method and equipment based on time series analysis, and the method and equipment solve the problems.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a power grid environment fire risk assessment method based on time sequence analysis, which mainly aims at realizing intelligent assessment of fire risk level in the current power grid environment.
The technical scheme for solving the technical problems is as follows:
a power grid environment fire risk assessment method based on time series analysis comprises the following steps:
s1, acquiring historical fire data in a power grid environment, and continuously acquiring power grid working parameters and environment monitoring data;
s2, data preprocessing;
s2.1, constructing a time sequence, and converting historical fire data, power grid working parameters and environment monitoring data into a format which can be analyzed by a machine learning model;
s2.2, carrying out data normalization and missing value processing on the constructed time sequence;
s3, classifying fire risks by adopting an unsupervised learning K-means clustering algorithm based on information recorded in a time sequence of historical fire data;
s4, selecting related time sequence features by adopting correlation analysis and principal component analysis, wherein independent variables of the time sequence are power grid working parameters and environment monitoring data;
s5, constructing a time sequence data set which is divided into training setsAnd test set->；
S6, constructing a machine learning model taking a convolutional neural network CNN and a long-short-term memory network LSTM as basic structures;
s7, collecting the test setAnd inputting the model into a machine learning model obtained through training, and carrying out model evaluation.
On the basis of the power grid environment fire risk assessment method based on machine learning time series analysis, the historical fire data comprises fire occurrence time, scale and duration, affected power grid facility type and scale, loss assessment and loss assessmentCasualties; the power grid working parameters comprise a voltage level U, a frequency f and a loadActive power P, reactive power Q, current I, voltage stability +.>Frequency stability->Line loss->Transformer capacity->And cable capacity->The method comprises the steps of carrying out a first treatment on the surface of the The environmental monitoring data includes air temperature->Wind speed->And atmospheric pressure->。
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific process of S2.1 is as follows:
(1) Data integration: aligning historical fire data, power grid working parameters and environment monitoring data according to the time stamp;
(2) Time sequence formatting: for each point in timeConstructing a containing time point +.>Data vector of all relevant data->；
(3) Sequence construction: arranging the data vectors in time sequence to form a time sequenceWherein->Is the length of the time series;
(4) Treatment time interval: the time interval of each data point in the time series is the same, and the time interval is uniformly set to 1 hour.
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific process of S2.2 is as follows:
(1) Data normalization: for each featureThe normalization process is expressed as:
,
wherein,is normalized data, ++>And->The minimum and maximum values of the feature in the whole dataset, respectively;
(2) Missing value processing:
interpolation is performed by the following steps:
a. adjacent data point selection: for the point in timeSelect adjacent known data points, assume +.>And->Time points +.>The last two known data points before and after;
b. linear interpolation: then, for the point in timeIs>The missing value is calculated by adopting linear interpolation, and the interpolation formula is as follows:
,
wherein,is the time point->Interpolation result of>And->Time points +.>And->Is a known characteristic value of (a).
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific implementation process is as follows:
(1) Determining a cluster number: elbow rule determines optimal cluster number +.>Calculate the difference +.>Total square error SSE at the value, point where SSE decrease is slowed down is selected as +.>Values, wherein SSE is calculated as follows:
,
wherein,is->Cluster (S)>Is a data point within a cluster,/>Is cluster->Centroid of->Is a dot->To centroid->Is the euclidean distance of (2);
(2) K-means clustering algorithm: randomly selectThe following procedure was repeated for each initial centroid until the centroid stabilized:
a. assigning each sample point to the nearest centroid, for each pointFind the closest centroid and assign:wherein->Is a dot->The cluster to which it is assigned;
b. re-computing the centroid of each cluster, the new centroid being the average of all points within the cluster:wherein->Is cluster->The number of points in (a);
(3) And (3) risk classification, namely marking each cluster as different risk classes according to a clustering result, so that the risk class of each cluster can be determined according to the characteristics of samples in the clusters.
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific implementation process is as follows:
(1) Correlation analysis: hypothesis testing features contained in the time series, for each pair of features By calculating their pearson correlation coefficients, the correlation between features is evaluated, the calculation process being:
,
wherein,and->Observations of the features, respectively +.>And->For their average values, setting a threshold value based on the result of the hypothesis test, pairs of features above which are considered highly correlated, randomly preserving one feature and removing the other in each pair of highly correlated features;
(2) And (3) principal component analysis: highly correlated feature pairs are identified and eliminated using principal component analysis PCA techniques to further reduce multiple collinearity between selected features, as follows:
a. calculating a covariance matrix: constructing covariance matrixes among features, and obtaining effective features and covariance matrixes by carrying out eigenvalue decomposition on the covariance matrixesElement->The calculation is as follows:
,
wherein,and->Are respectively->And->Characterised by->Normalized value on individual samples, +.>And->Is characterized by->And features->Average value after standardized treatment;
b. constructing a new feature set: performing feature decomposition on the covariance matrix, solving the feature value and the corresponding feature vector, arranging the feature vectors in descending order according to the feature value, and selecting the feature vectors beforeThe feature vectors are used as principal components to construct a new feature set using the selected principal components.
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific implementation process for constructing the time sequence data set is as follows:
by lengthI.e. 12 hours, to obtain +.>Time-series data, the constructed time-series data set is expressed as +.>Wherein->，Here, each->Is +.>Matrix of (2) < table->All characteristic sequences of bar data, each data point +.>Corresponds to a category label->Considering that each length is +>Contains fire risk rating sequence of corresponding length in the original data, and the data point +.>Fire risk rating class label corresponding to the same +.>Associated fire risk category label ++>Defined as the maximum risk level in the corresponding fire risk rating sequence, i.eWherein->Is->Fire risk level at moment;
time series data setWherein include->The pieces of time-series data are divided into two different subsets: training set->And test set->Comprising 80% and 20% time series data, respectively.
Based on the power grid environment fire risk assessment method based on machine learning time sequence analysis, the specific process of S6 is as follows:
s6.1 CNN feature extraction module:
the module utilizes CNN to extract spatial features in time sequence data, provides input for subsequent LSTM modules, and specifically, the CNN module is composed ofLayer convolution operation composition for +.>Layer (c):
input feature map dimension:the method comprises the steps of carrying out a first treatment on the surface of the Number of convolution kernels: />The method comprises the steps of carrying out a first treatment on the surface of the Convolution kernel size: /> The method comprises the steps of carrying out a first treatment on the surface of the Step size:the method comprises the steps of carrying out a first treatment on the surface of the Output feature map dimension: />,
Wherein the method comprises the steps of，/>And->The height, width and depth are calculated by the following formula: />；；/>，
Each layer of convolution is followed by a ReLU activation function and a max pooling layer;
first, theThe parameters of the layer pool layer are as follows: pooling window size: />Step size: />The pooled output dimensions are:
；
s6.2 LSTM time sequence processing module:
the LSTM module receives the characteristics extracted by the CNN module, and the LSTM unit of the module gradually processes the time sequence data and maintains an internal state, thereby effectively capturing the dynamic change and long-term dependence in the time sequence and outputting the moduleThe representation of the integrated time series dynamic characteristics, specifically, the number of hidden units of the LSTM layer is：
Input dimensions:output dimension: />For the LSTM time series processing module, the LSTM layer number is set to 1, LSTM unit number +.>Set to 100;
s6.3, a full connection layer classification module:
the full-connection layer classification module is used for converting the output of the LSTM module into a final classification result, and the full-connection layer receives the LSTM outputDimension feature vector, full connection layer with +.>Individual neurons, ->Category of classification for fire risk: input dimensions: />Output dimension: />；
The output layer converts the output of the neural network into a probability distribution using a softmax function:
wherein->And->Weight and bias, respectively, +.>Is the input of the full connection layer.
The second aspect of the invention also provides a computer device comprising a processor and a storage means adapted to store a plurality of program code adapted to be loaded and executed by the processor to perform a grid environment fire risk assessment method.
The third aspect of the present invention also provides a computer readable storage medium having stored therein a plurality of program code adapted to be loaded and executed by a processor to perform a grid environment fire risk assessment method.
The effects provided in the summary of the invention are merely effects of embodiments, not all effects of the invention, and the above technical solution has the following advantages or beneficial effects:
according to the invention, by integrating historical fire data, power grid parameters and environment monitoring data, constructing time series data and analyzing by using a machine learning method, comprehensive, accurate and real-time assessment of fire risks in the power grid environment is realized, the fire protection efficiency of the power grid environment is improved, and the fire protection effect in the power grid environment is enhanced.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
FIG. 1 is a schematic flow chart of the method of the present invention.
Detailed Description
In order to clearly illustrate the technical features of the present solution, the present invention will be described in detail below with reference to the following detailed description and the accompanying drawings. The following disclosure provides many different embodiments, or examples, for implementing different structures of the invention.
Example 1: a power grid environment fire risk assessment method based on time series analysis comprises the following steps:
s1, acquiring historical fire data in a power grid environment, and continuously acquiring power grid working parameters and environment monitoring data;
s2, data preprocessing;
s2.1, constructing a time sequence, and converting historical fire data, power grid working parameters and environment monitoring data into a format which can be analyzed by a machine learning model;
s2.2, carrying out data normalization and missing value processing on the constructed time sequence;
s3, classifying fire risks by adopting an unsupervised learning K-means clustering algorithm based on information recorded in a time sequence of historical fire data;
s4, selecting related time sequence features by adopting correlation analysis and principal component analysis, wherein independent variables of the time sequence are power grid working parameters and environment monitoring data;
s5, constructing a time sequence data set which is divided into training setsAnd test set->；
S6, constructing a machine learning model taking a convolutional neural network CNN and a long-short-term memory network LSTM as basic structures;
s7, collecting the test setAnd inputting the model into a machine learning model obtained through training, and carrying out model evaluation.
In this embodiment, the historical fire data includes fire occurrence time, scale, duration, type and scale of the affected grid facilities, loss assessment, and casualties; the power grid working parameters comprise a voltage level U, a frequency f and a loadActive power P, reactive power Q, current I, voltage stability +.>Frequency stability->Line lossTransformer capacity->And cable capacity->The method comprises the steps of carrying out a first treatment on the surface of the The environmental monitoring data includes air temperature->Wind speed->And atmospheric pressure->。
In this embodiment, the time series is a series of data point sets arranged in time sequence, which can reflect the trend and pattern of the data over time.
The specific process of time series generation is as follows:
(1) Data integration: aligning historical fire data, grid operating parameters and environmental monitoring data according to time stamps, which means that the data at each time point includes all relevant measurements and records;
(2) Time sequence formatting: for each point in timeConstruct a data vector +.>This vector contains +.>For example, grid parameters +.>And environmental parameters->Etc., wherein->、/>、/>Respectively->Grid parameters such as voltage, frequency, load at time, < >>、/>、/>；
(3) Sequence construction: arranging the data vectors in time sequence to form a time sequenceWherein->Is the length of the time series, +.>The characteristic value of each moment is the characteristic value;
(4) Treatment time interval: if the time intervals of data collection are inconsistent, interpolation or resampling is needed to ensure that the time intervals of each data point in the time series are the same, and the time intervals are uniformly set to be 1 hour in the embodiment.
After the time series is generated, in order to ensure the data quality and the effectiveness of model training, the data normalization and missing value processing operations are required, and the specific procedures are as follows:
(1) Data normalization: for each featureThe normalization process is expressed as:
,
wherein,is normalized data, ++>And->The minimum and maximum values of the feature in the whole dataset, respectively;
(2) Missing value processing:
in the actual data collection process, the situation of data missing is frequently encountered, for the missing values in the time series, the invention adopts an interpolation method, wherein the interpolation method is a method for deducing the value of an unknown data point according to known data points, is particularly suitable for filling the missing values of continuous time series data, and aims at each time point with the missing valuesInterpolation is performed by the following steps:
a. adjacent data point selection: for the point in timeSelect adjacent known data points, assume +.>And->Time points +.>The last two known data points before and after;
b. linear interpolation: for the point in timeIs>The missing value is calculated by adopting linear interpolation, and the interpolation formula is as follows:
,
wherein,is the time point->Interpolation result of>And->Time points +.>And->Is a known characteristic value of (a).
In this embodiment, based on the information recorded in the time series of the historical fire data, including the time, scale, duration, type and scale of the affected power grid facilities, loss assessment and casualties, the fire risk is classified by adopting an unsupervised learning K-means clustering algorithm, and the specific implementation process is as follows:
(1) Determining a cluster number: elbow rule determines optimal cluster number +.>Calculate the difference +.>Total square error SSE at the value, point where SSE decrease is slowed down is selected as +.>Values, wherein SSE is calculated as follows:
,
wherein,is->Cluster (S)>Is a data point within a cluster,/>Is cluster->Centroid of->Is a dot->To centroid->Is the euclidean distance of (2);
(2) K-means clustering algorithm: randomly selectThe following procedure was repeated for each initial centroid until the centroid stabilized:
a. assigning each sample point to the nearest centroid, for each pointFind the closest centroid and assign:wherein->Is a dot->The cluster to which it is assigned;
b. re-computing the centroid of each cluster, the new centroid being the average of all points within the cluster:wherein->Is cluster->The number of points in>Is->Centroid of each cluster;
(3) And (3) risk classification, namely marking each cluster as different risk classes according to a clustering result, so that the risk class of each cluster can be determined according to the characteristics of samples in the clusters.
In the power grid environment risk assessment method based on machine learning time sequence analysis, the independent variable is a time sequence formed by power grid working parameters and environment monitoring data, so as to eliminate irrelevant or redundant characteristics, facilitate the training of a subsequent machine learning model, select the related time sequence characteristics by adopting correlation analysis and principal component analysis, and specifically realize the following steps:
(1) Correlation analysis: hypothesis testing features contained in the time series, for each pair of features By calculating their pearson correlation coefficients, the correlation between features is evaluated, the calculation process being:
,
wherein,and->Observations of the features, respectively +.>And->For their average values, setting a threshold value based on the result of the hypothesis test, pairs of features above which are considered highly correlated, randomly preserving one feature and removing the other in each pair of highly correlated features;
(2) And (3) principal component analysis: highly correlated feature pairs are identified and eliminated using principal component analysis PCA techniques to further reduce multiple collinearity between selected features, as follows:
a. calculating a covariance matrix: constructing covariance matrixes among features, and obtaining effective features and covariance matrixes by carrying out eigenvalue decomposition on the covariance matrixesElement->The calculation is as follows:
,
wherein,and->Are respectively->And->Characterised by->Normalized value on individual samples, +.>And->Is characterized by->And features->Average value after standardized treatment;
b. constructing a new feature set: performing feature decomposition on the covariance matrix, solving the feature value and the corresponding feature vector, arranging the feature vectors in descending order according to the feature value, and selecting the feature vectors beforeThe feature vectors are used as principal components to construct a new feature set using the selected principal components.
In the present invention, S4 and S5 respectively generate fire risk ratings for each momentAnd related feature vector->Wherein the dimension of the feature vector ∈ ->And the total length of the time series is +.>The specific implementation process for constructing the time sequence data set is as follows:
by lengthI.e. 12 hours, to obtain +.>Time-series data, the constructed time-series data set is expressed as +.>Wherein->，Here, each->Is +.>Is a matrix of->All characteristic sequences of bar data, each data point +.>Corresponds to a category label->Considering that each length is +>Contains fire risk rating sequence of corresponding length in the original data, and the data point +.>Fire risk rating class label corresponding to the same +.>Associated fire risk category label ++>Defined as the maximum risk level in the corresponding fire risk rating sequence, i.eWherein->Is->Fire risk level at moment;
time series data setWherein include->The pieces of time-series data are divided into two different subsets: training set->And test set->Comprising 80% and 20% time series data, respectively.
The invention builds a machine learning model with a convolutional neural network CNN and a long-short-term memory network LSTM as basic structures, and the model combines the spatial feature extraction capacity of the CNN and the time sequence data processing capacity of the LSTM, and specifically comprises the following modules:
s6.1 CNN feature extraction module:
the module utilizes CNN to extract spatial features in time sequence data, provides input for subsequent LSTM modules, and specifically, the CNN module is composed ofLayer convolution operation composition for +.>Layer (c):
input feature map dimension:the method comprises the steps of carrying out a first treatment on the surface of the Number of convolution kernels: />The method comprises the steps of carrying out a first treatment on the surface of the Convolution kernel size: /> The method comprises the steps of carrying out a first treatment on the surface of the Step size:the method comprises the steps of carrying out a first treatment on the surface of the Output feature map dimension: />,
Wherein the method comprises the steps of，/>And->The height, width and depth are calculated by the following formula: />；；/>，
Each layer of convolution is followed by a ReLU activation function and a max pooling layer;
first, theThe parameters of the layer pool layer are as follows: pooling window size: />Step size: />The pooled output dimensions are:
；
s6.2 LSTM time sequence processing module:
the LSTM module receives the features extracted by the CNN module, the LSTM unit of the module gradually processes the time sequence data and maintains an internal state, thereby effectively capturing dynamic changes and long-term dependence in the time sequence, the module output is a representation integrating the dynamic features of the time sequence, and specifically, the number of hidden units of the LSTM layer is as follows：
Input dimensions:output dimension: />For the LSTM time series processing module, the LSTM layer number is set to 1, LSTM unit number +.>Set to 100;
s6.3, a full connection layer classification module:
the full-connection layer classification module is used for converting the output of the LSTM module into a final classification result, and the full-connection layer receives the LSTM outputDimension feature vector, full connection layer with +.>Individual neurons, ->Category of classification for fire risk: input dimensions: />Output dimension: />；
The output layer converts the output of the neural network into a probability distribution using a softmax function:
wherein->And->Weight and bias, respectively, +.>Is the input of the full connection layer.
For fire risk assessment, it is common practice to model such prediction problems as classification problems based on sample features, such as support vector machines SVM, decision trees, and deep neural networks. From the perspective of time sequence analysis, the risk information implied by the change of the power grid parameters and the environment monitoring data before and after the occurrence of the fire disaster in the power grid environment is fully excavated, and the accuracy of fire risk assessment can be further improved. The research utilizes historical fire data, power grid parameters and environment monitoring data generated in a simulation environment to respectively establish a prediction model based on traditional sample characteristics and a prediction model based on time sequence analysis. As shown in table 1, the prediction effect of the present invention and each model on the risk of fire is known, and compared with the prior art, the technology provided by the present invention has higher accuracy.
Table 1 effect of the invention and models on prediction of fire risk:
,
example 2: a computer device comprising a processor and storage means, said storage means being adapted to store a plurality of program code,
wherein said program code is adapted to be loaded and executed by said processor to perform said grid environment fire risk assessment method.
Example 3: a computer readable storage medium having stored therein a plurality of program codes adapted to be loaded and executed by a processor to perform the grid environment fire risk assessment method.
While the foregoing description of the embodiments of the present invention has been presented with reference to the drawings, it is not intended to limit the scope of the invention, but rather, it is apparent that various modifications or variations can be made by those skilled in the art without the need for inventive work on the basis of the technical solutions of the present invention.
Claims (10)
1. A power grid environment fire risk assessment method based on time sequence analysis is characterized by comprising the following steps: the method comprises the following steps:
s1, acquiring historical fire data in a power grid environment, and continuously acquiring power grid working parameters and environment monitoring data;
s2, data preprocessing;
s2.1, constructing a time sequence, and converting historical fire data, power grid working parameters and environment monitoring data into a format which can be analyzed by a machine learning model;
s2.2, carrying out data normalization and missing value processing on the constructed time sequence;
s3, classifying fire risks by adopting an unsupervised learning K-means clustering algorithm based on information recorded in a time sequence of historical fire data;
s4, selecting related time sequence features by adopting correlation analysis and principal component analysis, wherein independent variables of the time sequence are power grid working parameters and environment monitoring data;
s5, constructing a time sequence data set which is divided into training setsAnd test set->；
S6, constructing a machine learning model taking a convolutional neural network CNN and a long-short-term memory network LSTM as basic structures;
s7, collecting the test setAnd inputting the model into a machine learning model obtained through training, and carrying out model evaluation.
2. The method for evaluating the fire risk of the power grid environment based on time series analysis according to claim 1, wherein the method comprises the following steps of: the historical fire data comprises fire occurrence time, scale, duration, affected power grid facility type and scale, loss assessment and casualties; the power grid working parameters comprise a voltage level U, a frequency f and a loadActive power P, reactive power Q, current I, voltage stability +.>Frequency stability->Line loss->Capacity of transformerAnd cable capacity->The method comprises the steps of carrying out a first treatment on the surface of the The environmental monitoring data includes air temperature->Humidity->Wind speed->And atmospheric pressure->。
3. The method for evaluating the fire risk of the power grid environment based on time series analysis according to claim 2, wherein the method is characterized by comprising the following steps of: s2.1 specifically comprises the following steps:
(1) Data integration: aligning historical fire data, power grid working parameters and environment monitoring data according to the time stamp;
(2) Time sequence formatting: for each point in timeConstructing a containing time point +.>Data vector of all relevant data->；
(3) Sequence construction: arranging the data vectors in time sequence to form a time sequenceWhereinIs the length of the time series;
(4) Treatment time interval: the time interval of each data point in the time series is the same, and the time interval is uniformly set to 1 hour.
4. The method for evaluating the fire risk of the power grid environment based on time series analysis according to claim 1, wherein the method comprises the following steps of: s2.2, the specific process is as follows:
(1) Data normalization: for each featureThe normalization process is expressed as:
，
wherein,is normalized data, ++>And->The minimum and maximum values of the feature in the whole dataset, respectively;
(2) Missing value processing:
interpolation is performed by the following steps:
a. adjacent data point selection: for the point in timeSelect adjacent known data points, assume +.>And->Time points +.>The last two known data points before and after;
b. linear interpolation: then, for the point in timeIs>The missing value is calculated by adopting linear interpolation, and the interpolation formula is as follows:
，
wherein,is the time point->Interpolation result of>And->Time points +.>And->Is a known characteristic value of (a).
5. The power grid environment fire risk assessment method based on time series analysis according to claim 1, wherein the specific implementation process is as follows:
(1) Determining a cluster number: elbow rule determines optimal cluster number +.>Calculate the difference +.>Total square error SSE at the value, point where SSE decrease is slowed down is selected as +.>Values, wherein SSE is calculated as follows:
，
wherein,is->Cluster (S)>Is a data point within a cluster,/>Is cluster->Centroid of->Is a dot->To centroid->Is the euclidean distance of (2);
(2) K-means clustering algorithm: randomly selectThe following procedure was repeated for each initial centroid until the centroid stabilized:
a. assigning each sample point to the nearest centroid, for each pointFind the closest centroid and assign:wherein->Is a dot->The cluster to which it is assigned;
b. re-computing the centroid of each cluster, the new centroid being the average of all points within the cluster:wherein->Is cluster->The number of points in (a);
(3) And (3) risk classification, namely marking each cluster as different risk classes according to a clustering result, so that the risk class of each cluster can be determined according to the characteristics of samples in the clusters.
6. The power grid environment fire risk assessment method based on time series analysis according to claim 1, wherein the specific implementation process is as follows:
(1) Correlation analysis: hypothesis testing features contained in the time series, for each pair of featuresBy calculating their pearson correlation coefficients, the correlation between features is evaluated, the calculation process being:
,
wherein,and->Observations of the features, respectively +.>And->For their average values, setting a threshold value based on the result of the hypothesis test, pairs of features above which are considered highly correlated, randomly preserving one feature and removing the other in each pair of highly correlated features;
(2) And (3) principal component analysis: highly correlated feature pairs are identified and eliminated using principal component analysis PCA techniques to further reduce multiple collinearity between selected features, as follows:
a. calculating a covariance matrix: constructing covariance matrixes among features, and obtaining effective features and covariance by performing eigenvalue decomposition on the covariance matrixesVariance matrixElement->The calculation is as follows:
,
wherein,and->Are respectively->And->Characterised by->Normalized value on individual samples, +.>And->Is characterized by->And features->Average value after standardized treatment;
b. constructing a new feature set: performing feature decomposition on the covariance matrix, solving the feature value and the corresponding feature vector, arranging the feature vectors in descending order according to the feature value, and selecting the feature vectors beforeThe feature vectors are used as principal components to construct a new feature set using the selected principal components.
7. The power grid environment fire risk assessment method based on time series analysis according to claim 1, wherein the specific implementation process of constructing the time series data set is as follows:
by lengthI.e. 12 hours, to obtain +.>Time-series data, the constructed time-series data set is expressed as +.>Wherein->，/>Here, each->Is +.>Is a matrix of->All characteristic sequences of bar data, each data point +.>Corresponds to a category label->Considering that each length is +>Contains fire risk rating sequence of corresponding length in the original data, and the data point +.>Fire risk rating class label corresponding to the same +.>Associated fire risk category label ++>Defined as the maximum risk level in the corresponding fire risk rating sequence, i.e. +.>Wherein->Is->Fire risk level at moment;
time series data setWherein include->The pieces of time-series data are divided into two different subsets: training set->And test set->Comprising 80% and 20% time series data, respectively.
8. The power grid environment fire risk assessment method based on time series analysis according to claim 1, wherein the specific process of S6 is as follows:
s6.1 CNN feature extraction module:
the module utilizes CNN to extract spatial features in time sequence data, provides input for subsequent LSTM modules, and specifically, the CNN module is composed ofLayer convolution operation composition for +.>Layer (c):
input feature map dimension:number of convolution kernels: />The method comprises the steps of carrying out a first treatment on the surface of the Convolution kernel size: /> The method comprises the steps of carrying out a first treatment on the surface of the Step size: />The method comprises the steps of carrying out a first treatment on the surface of the Output feature map dimension: />,
Wherein the method comprises the steps of，/>And->The height, width and depth are calculated by the following formula:
,
each layer of convolution is followed by a ReLU activation function and a max pooling layer;
first, theThe parameters of the layer pool layer are as follows: pooling window size: />Step size: />The pooled output dimensions are:
,
s6.2 LSTM time sequence processing module:
the LSTM module receives the features extracted by the CNN module, the LSTM unit of the module gradually processes the time sequence data and maintains an internal state, thereby effectively capturing dynamic changes and long-term dependence in the time sequence, the module output is a representation integrating the dynamic features of the time sequence, and specifically, the number of hidden units of the LSTM layer is as follows：
Input dimensions:output dimension: />For the LSTM time series processing module,the number of LSTM layers is set to 1, the number of LSTM cells +.>Set to 100;
s6.3, a full connection layer classification module: the full-connection layer classification module is used for converting the output of the LSTM module into a final classification result, and the full-connection layer receives the LSTM outputDimension feature vector, full connection layer with +.>Individual neurons, ->Category of classification for fire risk: input dimensions: />Output dimension: />；
The output layer converts the output of the neural network into a probability distribution using a softmax function:
wherein->And->Weight and bias, respectively, +.>Is the input of the full connection layer.
9. A computer device comprising a processor and storage means, said storage means being adapted to store a plurality of program code,
characterized in that the program code is adapted to be loaded and run by the processor to perform the grid environment fire risk assessment method of any one of claims 1 to 8.
10. A computer readable storage medium, in which a plurality of program codes are stored, characterized in that the program codes are adapted to be loaded and run by a processor to perform the grid environment fire risk assessment method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410070578.8A CN117592789B (en) | 2024-01-18 | 2024-01-18 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410070578.8A CN117592789B (en) | 2024-01-18 | 2024-01-18 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117592789A true CN117592789A (en) | 2024-02-23 |
| CN117592789B CN117592789B (en) | 2024-04-16 |
Family
ID=89913672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410070578.8A Active CN117592789B (en) | 2024-01-18 | 2024-01-18 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117592789B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360369A (en) * | 2018-09-19 | 2019-02-19 | 李琳 | One kind carrying out fire hazard analysis method and device based on cluster result |
| CN109472421A (en) * | 2018-11-22 | 2019-03-15 | 广东电网有限责任公司 | A kind of power grid mountain fire sprawling method for early warning and device |
| CN109686036A (en) * | 2019-01-09 | 2019-04-26 | 深圳市中电数通智慧安全科技股份有限公司 | A kind of fire monitoring method, device and edge calculations device |
| CN112308292A (en) * | 2020-10-10 | 2021-02-02 | 广东工业大学 | Method for drawing fire risk grade distribution map |
| WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
| CN112785066A (en) * | 2021-01-27 | 2021-05-11 | 北京师范大学 | Convolution-recurrent neural network-based global wildfire sub-season time-space prediction method |
| CN113177737A (en) * | 2021-05-26 | 2021-07-27 | 南京恩瑞特实业有限公司 | Urban rainstorm disaster risk assessment method and system based on GA (genetic algorithm) optimization BP (back propagation) neural network |
| CN113553764A (en) * | 2021-07-13 | 2021-10-26 | 广东工业大学 | Mountain fire prediction method based on deep learning network |
| US20230074074A1 (en) * | 2021-09-02 | 2023-03-09 | Southwest Petroleum University | Intelligent recognition method for while-drilling safety risk based on convolutional neural network |
| KR20230102532A (en) * | 2021-12-30 | 2023-07-07 | 주식회사 현대케피코 | Fire detection system and method using time series data prediction thchnique and image classification |
| CN116843074A (en) * | 2023-07-06 | 2023-10-03 | 南宁师范大学 | Typhoon disaster damage prediction method based on CNN-LSTM model |
| CN117152893A (en) * | 2023-10-31 | 2023-12-01 | 广州市林业和园林科学研究院 | Forest disaster prevention method and system |
-
2024
- 2024-01-18 CN CN202410070578.8A patent/CN117592789B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360369A (en) * | 2018-09-19 | 2019-02-19 | 李琳 | One kind carrying out fire hazard analysis method and device based on cluster result |
| CN109472421A (en) * | 2018-11-22 | 2019-03-15 | 广东电网有限责任公司 | A kind of power grid mountain fire sprawling method for early warning and device |
| CN109686036A (en) * | 2019-01-09 | 2019-04-26 | 深圳市中电数通智慧安全科技股份有限公司 | A kind of fire monitoring method, device and edge calculations device |
| WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
| CN112308292A (en) * | 2020-10-10 | 2021-02-02 | 广东工业大学 | Method for drawing fire risk grade distribution map |
| CN112785066A (en) * | 2021-01-27 | 2021-05-11 | 北京师范大学 | Convolution-recurrent neural network-based global wildfire sub-season time-space prediction method |
| CN113177737A (en) * | 2021-05-26 | 2021-07-27 | 南京恩瑞特实业有限公司 | Urban rainstorm disaster risk assessment method and system based on GA (genetic algorithm) optimization BP (back propagation) neural network |
| CN113553764A (en) * | 2021-07-13 | 2021-10-26 | 广东工业大学 | Mountain fire prediction method based on deep learning network |
| US20230074074A1 (en) * | 2021-09-02 | 2023-03-09 | Southwest Petroleum University | Intelligent recognition method for while-drilling safety risk based on convolutional neural network |
| KR20230102532A (en) * | 2021-12-30 | 2023-07-07 | 주식회사 현대케피코 | Fire detection system and method using time series data prediction thchnique and image classification |
| CN116843074A (en) * | 2023-07-06 | 2023-10-03 | 南宁师范大学 | Typhoon disaster damage prediction method based on CNN-LSTM model |
| CN117152893A (en) * | 2023-10-31 | 2023-12-01 | 广州市林业和园林科学研究院 | Forest disaster prevention method and system |
Non-Patent Citations (4)
| Title |
|---|
| GUIDO R 等: "Global fire emissions estimates during 1997–2016", EARTH SYST. SCI. DATA, 31 December 2017 (2017-12-31), pages 697 - 720 * |
| 杨淳岚 等: "电网山火监测及风险预警技术研究综述", 电网技术, vol. 47, no. 11, 30 November 2023 (2023-11-30), pages 4765 - 4777 * |
| 翁国庆;龚阳光;舒俊鹏;黄飞腾;: "基于聚类LSTM深度学习模型的主动配电网电能质量预测", 高技术通讯, no. 07, 15 July 2020 (2020-07-15), pages 687 - 697 * |
| 陆继翔 等: "基于CNN-LSTM混合神经网路模型的短期负荷预测方法", 电力系统自动化, vol. 43, no. 8, 31 December 2019 (2019-12-31), pages 131 - 137 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117592789B (en) | 2024-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115276006B (en) | Load prediction method and system for power integration system | |
| CN112382352B (en) | Method for quickly evaluating structural characteristics of metal organic framework material based on machine learning | |
| Lu et al. | Dominant feature selection for the fault diagnosis of rotary machines using modified genetic algorithm and empirical mode decomposition | |
| CN113570138B (en) | Method and device for predicting residual service life of equipment of time convolution network | |
| CN115021679B (en) | Photovoltaic equipment fault detection method based on multi-dimensional outlier detection | |
| CN111784061B (en) | Training method, device and equipment for power grid engineering cost prediction model | |
| CN116737510B (en) | Data analysis-based intelligent keyboard monitoring method and system | |
| CN111639722A (en) | Transformer fault diagnosis method based on principal component analysis and twin support vector machine | |
| CN113159225B (en) | Multivariable industrial process fault classification method | |
| CN113110961B (en) | Equipment abnormality detection method and device, computer equipment and readable storage medium | |
| CN114357870A (en) | Metering equipment operation performance prediction analysis method based on local weighted partial least squares | |
| CN117313015A (en) | Time sequence abnormality detection method and system based on time sequence and multiple variables | |
| CN113484817A (en) | Intelligent electric energy meter automatic verification system abnormity detection method based on TSVM model | |
| CN117520809A (en) | Transformer fault diagnosis method based on EEMD-KPCA-CNN-BiLSTM | |
| CN117592789B (en) | Power grid environment fire risk assessment method and equipment based on time sequence analysis | |
| CN116400266A (en) | Transformer fault detection method, device and medium based on digital twin model | |
| CN116400168A (en) | Power grid fault diagnosis method and system based on depth feature clustering | |
| CN115659271A (en) | Sensor abnormality detection method, model training method, system, device, and medium | |
| CN114648076A (en) | Unsupervised learning battery production process abnormal fluctuation detection method | |
| CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
| Singh et al. | Predicting the remaining useful life of ball bearing under dynamic loading using supervised learning | |
| CN110543869A (en) | Ball screw service life prediction method and device, computer equipment and storage medium | |
| CN117131425B (en) | Numerical control machine tool processing state monitoring method and system based on feedback data | |
| CN112685933B (en) | Method for predicting residual service life of roller screw pair | |
| CN117556753B (en) | Method, device, equipment and storage medium for analyzing energy consumption of storage chip |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |