CN117036103A - Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network - Google Patents
Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network Download PDFInfo
- Publication number
- CN117036103A CN117036103A CN202311057728.3A CN202311057728A CN117036103A CN 117036103 A CN117036103 A CN 117036103A CN 202311057728 A CN202311057728 A CN 202311057728A CN 117036103 A CN117036103 A CN 117036103A
- Authority
- CN
- China
- Prior art keywords
- data
- power system
- lstm
- neural network
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 17
- 238000003062 neural network model Methods 0.000 claims abstract description 10
- 238000012360 testing method Methods 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 16
- 230000000306 recurrent effect Effects 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 7
- 125000004122 cyclic group Chemical group 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 3
- 238000012502 risk assessment Methods 0.000 abstract description 2
- 210000004027 cell Anatomy 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000013095 identification testing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- 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
- G06N3/08—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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Physics (AREA)
- Development Economics (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
- Public Health (AREA)
- Educational Administration (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses an LSTM (least squares) circulating neural network-based power system operation situation prediction method. Secondly, building a situation prediction model of the power system, wherein an operation state identification module and an operation data trend prediction module contained in the model are built by an LSTM circulating neural network model. The running state recognition module and the running data trend prediction module are trained and tested by using the respective samples. The risk assessment is carried out on the running state of the whole power system by identifying the running state of each line and predicting the running data trend.
Description
Technical Field
The invention relates to the field of situation awareness research of power systems, in particular to a power system operation situation prediction method based on an LSTM (least squares) circulating neural network.
Background
With the continuous development of economy and the continuous increase of power demand, the scale of a power system is continuously enlarged, and the running condition of a power grid is increasingly complex. Therefore, the power construction task is very difficult, and it is important to ensure the safety and reliability of the operation of the power system. The power system has the characteristics of large scale, multiple devices, wide coverage area, long operation time and the like, comprises links of power generation, power transmission, power transformation, power distribution and the like, and can generate a large amount of operation detection data in the operation process of the whole power system. With the rapid development of computer information technology and storage technology, the technology of power grid digitization and informatization is mature, and the power system enters a big data age. This makes possible a deep learning method based on the power system operation data. With the improvement of hardware computing capability, deep learning technology represented by neural networks is rapidly developed, and is widely applied to the aspects of power system operation monitoring, fault detection, load prediction and the like.
In recent years, research on power system problems has gradually been converted from a conventional artificial intelligence method to a deep learning method based on power big data. The deep learning method brings new breakthrough to the problem of power system faults for the data mining capability, the nonlinear characterization capability and the strong robustness. However, in existing deep learning-based power system situation awareness studies, there are major limitations, firstly, to the power distribution network for operational state identification and operational data prediction, and secondly, to the short-term power load or other single attribute data. In order to improve the accuracy of the operation situation prediction of the power system, the operation state of each power transmission line is identified, and meanwhile, the operation trend of key data in the lines is predicted as much as possible. By analyzing the running states of all lines in the system and the running trend of key attribute data, the running states of the whole system are accurately estimated and predicted.
In general, a deep learning method based on data mining has been widely used in the research of power systems, but there is still a limitation in the potential perception of power system states. In order to more accurately predict the operation situation of the power system, the operation state of each power transmission line needs to be identified and the operation trend of key attribute data of the line itself needs to be predicted.
Disclosure of Invention
Based on the deficiency of the existing power system situation prediction strategy, in order to solve the problems existing in the background technology, the invention provides a power system running situation prediction method based on an LSTM circulating neural network.
The invention particularly relates to a power system operation situation prediction model constructed based on historical operation data of a power transmission line and an LSTM cyclic neural network model. The system situation prediction model mainly comprises two modules, namely an operation state identification module and an operation data trend prediction module, wherein the operation state identification module and the operation data trend prediction module are respectively used for real-time identification of an operation state and prediction of an operation trend of key data. Firstly, after a line is selected, the historical operation data of the line is processed to generate a corresponding state identification sample and a data trend prediction sample, the state identification sample is analyzed by an operation state identification module to judge the operation state of the line, if the line is in a fault state, a specific fault type is output, and if the line is in a normal operation state, the line operation data trend is predicted. By analyzing the running states of all lines and the running trend of key attribute data in the power system, the running states of the whole system are further accurately estimated and predicted, so that the faults of the power system can be prevented to a certain extent, the safe and reliable running of the power grid is ensured, countermeasures can be timely made after the system faults, the power supply is recovered at the fastest speed, and the loss is reduced to the greatest extent.
In order to solve the technical problems, the technical scheme of the invention is as follows:
an LSTM (least squares) circulating neural network-based power system operation situation prediction method comprises the following steps:
s1, processing power line operation data and generating samples
S1-1, carrying out normalization processing on an initial data sample composed of power line operation data.
S1-2, performing sliding sampling on the initial data sample after normalization processing to generate a final data sample.
S1-3, performing outlier processing on the final data sample.
S2, building a situation prediction model of the power system
S2-1, a power system situation prediction model comprises two modules: the running state identification module and the running data trend prediction module are both built by an LSTM circulating neural network model.
S3, training a power system situation prediction model through the sample obtained in the S1
S3-1, randomly scrambling the data samples processed in the S1, and dividing the data samples into a training set, a verification set and a test set according to the proportion.
S3-2, the running state identification module adjusts the sliding time window depth and the network structure super-parameters according to the identification accuracy of the test set.
S3-3, the operation data trend prediction module adjusts the sliding time window and the network structure super-parameters according to the error size of the predicted value.
S3-4, training by using respective samples by the running state identification module and the running data trend prediction module.
S4, a test running state identification module and a running data trend prediction module acquire a power system running state prediction result.
Preferably, in the step S1-1, for any line, two ends of the line are respectively labeled as an I end and a J end. Specifically, the active power of the I end refers to the active power flowing through the I end, and the direction of the active power is from the I end to the J end; the active power at the J end refers to the active power flowing through the J end, and the direction from the J end to the I end. The initial data sample is composed of operation data of a power line and mainly comprises 10 kinds of attribute data such as I-terminal active power, I-terminal reactive power, J-terminal active power, J-terminal reactive power, an I-terminal disconnection mark, a J-terminal disconnection mark and the like, and the sampling time interval is 1 hour.
The running state identification module uses all 10 kinds of attribute data to train, verify and test; the operation data trend prediction module uses 4 key attribute data including I-end active power, I-end reactive power, J-end active power and J-end reactive power to train, verify and test.
Preferably, in the step S1-3, the outlier of the sample includes a block of zero, non-numeric value (NAN value), and the final data sample with the outlier is screened out.
Preferably, in the step S2-1, the operating state recognition module and the operating data trend prediction module use different time window depths, and the sliding window of the operating data trend prediction module is far deeper than the operating state recognition module.
Preferably, in the step S2-1, the operation state identification module accurately identifies the operation state of the current line according to the operation data. The operation data trend prediction module predicts future time operation data according to the historical operation data. The running state identification module and the running data trend prediction module are both composed of an LSTM cyclic neural network model, and the structure of the running data trend prediction module is more complex than that of the running state identification module.
The invention designs an LSTM circulating neural network-based power system operation situation prediction method. The system mainly comprises two modules, wherein the first module is an operation state identification module, a state identification sample is generated by processing line operation data, and the operation state of each line is identified and classified according to the generated state identification sample. The second module is an operation data trend prediction module, and after the line is judged to be in a normal operation state by the operation state identification module, a data trend prediction sample is generated by processing operation data of the normal operation line, and the operation data trend of each line is predicted according to the generated data trend prediction sample.
The invention relates to a method for predicting the running situation of a power system. The invention aims to accurately predict the running situation of the power system through historical running data of the power system and an LSTM cyclic neural network model.
The invention has the following characteristics and beneficial effects:
by adopting the technical scheme, the historical operation data of the power system is processed to generate the corresponding state identification training sample and the data trend prediction training sample, and the corresponding state identification training sample and the data trend prediction training sample are respectively input into the built operation state identification module and the operation data trend prediction module to train, so that the prediction of the operation situation of the power system is realized. According to the invention, on one hand, the operation state identification module is trained based on historical operation data, so that the operation state of each line of the system is identified by the operation data, and on the other hand, the operation data trend prediction module is trained based on the historical operation data, so that the operation trend of key data of each line in the system is predicted by the operated data. Through the identification of the running states of each line and the prediction of the running data trend, the risk assessment can be further carried out on the running states of the whole power system, the fault of the power system can be prevented to a certain extent, the safe and reliable running of the power grid is ensured, countermeasures can be timely carried out after the system fault, the power supply is recovered at the fastest speed, and the loss is reduced to the greatest extent.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
FIG. 1 is a flow chart of a power system operation situation prediction method based on an LSTM cyclic neural network in an embodiment of the invention;
FIG. 2 is a schematic diagram of an LSTM cell according to an embodiment of the invention;
FIG. 3 is a schematic diagram of an LSTM recurrent neural network model according to an embodiment of the invention;
FIG. 4 is a confusion matrix tested by a power system line operation status recognition module according to an embodiment of the present invention;
FIG. 5 is a graph showing the effect of the trend prediction module of the line operation data of the power system on predicting the active power of the I end according to the embodiment of the invention;
FIG. 6 is a graph showing the effect of the power system line operation data trend prediction module on predicting reactive power at the I end in an embodiment of the present invention;
FIG. 7 is a graph showing the effect of the trend prediction module of the line operation data of the power system on predicting the active power of the J end according to the embodiment of the present invention;
fig. 8 is a graph of the effect of the power system line operation data trend prediction module on the prediction of the reactive power at the J end according to the embodiment of the present invention.
Detailed Description
It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other.
In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first", "a second", etc. may explicitly or implicitly include one or more such feature. In the description of the present invention, unless otherwise indicated, the meaning of "a plurality" is two or more.
In the description of the present invention, it should be noted that, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art in a specific case.
The invention provides a power system running situation prediction method based on an LSTM circulating neural network, which is shown in figure 1 and comprises the following steps:
s1, processing data and generating samples
S1-1, carrying out normalization processing on an initial data sample composed of power line operation data.
In particular, the initial data samples used mainly comprise two types, one being initial samples in a normal operating state and the other being initial samples in a corresponding fault state. For any power line, the two ends of the power line are respectively marked as an I end and a J end, and the two types of samples mainly comprise I end active power, I end reactive power, J end active power, J end reactive power, an I end disconnection mark, a J end disconnection mark, an I end calculation node, a J end calculation node, an I end electric island and a J end electric island, which are ten kinds of attribute data. During normalization processing, except the I end disconnection mark and the J end disconnection mark, all the rest data are normalized according to each attribute column,
wherein X is normal For normalized data, X is the data to be normalized, X max 、X min Respectively, maximum value and minimum value in the collected single attribute data.
S1-2, performing sliding sampling on the initial sample to generate a final data sample.
It should be noted that, for the running state identification module, a sliding window with a sliding window depth of 5 is selected to perform sliding sampling on all ten columns of attribute data of the initial sample to generate a training sample, and the label is the running state of the line corresponding to the initial sample, where the label contains running state labels in 5 of normal running, single-phase fault, two-way fault, three-phase fault and other faults. For the running data trend prediction module, a sliding window with the depth of 168 is selected, four attribute data of the I-end active power, the I-end reactive power, the J-end active power and the J-end reactive power are subjected to sliding sampling to generate a training sample, and the corresponding predicted data label is four attribute data values at the next moment at the bottom of the window.
S1-3, performing outlier processing on the final data sample.
The training samples generated after the sliding acquisition had the problems of zero blocking and NAN value, and samples having abnormal value problems were detected and removed.
S2, building a situation prediction model of the power system
S2-1, a power system situation prediction model comprises two modules, namely an operation state identification module and an operation data trend prediction module, which are built by an LSTM circulating neural network model.
It should be noted that, a plurality of LSTM units are used to form an LSTM layer, and the LSTM cyclic neural network model is built through the combination of the LSTM layers and the full connection layer. The LSTM network model structure is shown in FIG. 3, and mainly comprises an LSTM layer and a full connection layer. The LSTM layer is mainly used for extracting characteristics of sample data, and the full-connection layer is used for mapping the extracted characteristic data to a sample marking space, so that classification and regression tasks are realized. The input sample is subjected to extraction of data characteristic information through two LSTM layers, and then the extracted characteristic information is sent to a fully-connected network for classification or prediction.
As shown in fig. 2, the LSTM recurrent neural network model is composed of an LSTM layer and a fully connected layer. The LSTM layer is mainly composed of LSTM cells for extraction of sample data features, while the full connection layer functions to map the extracted feature data to a sample marker space. The LSTM unit comprises a cellular state (C t ) Forget door (f) t ) Input door (i) t ) And output door (o) t ) The transmission state is controlled by the gating state, the important information is selectively memorized,noise information is filtered, and time series data can be effectively learned. Wherein x is t F is the data of the present time t I is a forgetful door t O is an input door t In order to output the door, the door is provided with a door opening,c is the current input state t For the current cell state, C t-1 H is the state of the last time cell t For the output of the instant unit, h t-1 Is the output of the last cell. Will x at each time instant processing t And h t-1 The two matrices are directly stitched together as input, plus the weight ω and the bias b. Forget door to last cell state C t-1 And (5) carrying out selective forgetting, wherein the obtained output is the memory of the history information. Input gate is for the current input state->And selectively inputting, wherein the obtained output is the input information. The memory of the history information and the summation of the information input at this time are the state C of the unit at this time t . The output gate selectively outputs the current unit state to obtain a final output h t 。
f t ＝σ(ω f ·[h (t-1) ，x t ]+b f )
i t ＝σ(ω i ·[h (t-1) ，x t ]+b i )
o t ＝σ(ω o ·[h (t-1) ，x t ]+b o )
h t ＝o t ·tanh(C t )
The sample is subjected to extraction of data characteristic information through two LSTM layers, and then the extracted characteristic information is sent to a fully-connected network for classification or prediction.
S3, training a power system situation prediction model through the sample obtained in the S1
S3-1, randomly scrambling the data samples processed in the S1, and dividing the data samples into a training set, a verification set and a test set according to the proportion.
S3-2, the running state identification module adjusts the sliding time window depth and the network structure super-parameters according to the identification accuracy of the test set. The depth of the selected time window is adjusted to be 5, the network model comprises two LSTM layers and two full-connection layers, wherein the number of LSTM units of the LSTM layers is 128 for the first layer and 256 for the second layer, and the number of neurons of the full-connection layers is 5 for the first layer and 128 for the second layer.
S3-3, the running data trend prediction module adjusts the sliding time window and the network structure super-parameters according to the error of the predicted value. The depth of the selected time window is adjusted to 168, the network model comprises two LSTM layers and two full-connection layers, wherein the number of LSTM units of the LSTM layers is 256 in the first layer and 512 in the second layer, and the number of neurons of the full-connection layers is 4 in the first layer and 256 in the second layer.
S3-4, training by using respective samples by the running state identification module and the running data trend prediction module.
S4, testing performance of the running state identification module and the running data trend prediction module.
Specifically, the trained model is tested by using the test set, and the running state recognition module tests the confusion matrix as shown in fig. 4. As can be seen from fig. 4, the operation state identification module performs identification test on five states of normal operation, single-phase fault, two-phase fault, three-phase fault and indeterminate fault, the accuracy is 100%, 94.84%, 97.22%, 93.08% and 95.34%, and the overall average accuracy of the model is 96.58%. The test results of the operation data trend prediction module are shown in fig. 5, 6, 7 and 8, and are respectively I-terminal active power prediction, I-terminal reactive power prediction, J-terminal active power prediction and J-terminal reactive power prediction. In order to accurately evaluate the model, 100 complete data samples are randomly selected for testing, and the prediction errors of the I-end active power, the I-end reactive power, the J-end active power and the J-end reactive power are calculated to be 5.28%, 7.67%, 5.47% and 8.41% respectively.
The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It will be apparent to those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments, including the components, without departing from the principles and spirit of the invention, yet fall within the scope of the invention.
Claims (7)
1. The power system operation situation prediction method based on the LSTM circulating neural network is characterized by comprising the following steps of:
s1, processing power line operation data and generating a sample;
s2, building a situation prediction model of the power system, wherein the prediction model comprises two modules: the running state identification module and the running data trend prediction module are built by an LSTM cyclic neural network model;
s3, training a situation prediction model of the power system through the sample obtained in the S1;
s4, a test running state identification module and a running data trend prediction module acquire a power system running state prediction result.
2. The method for predicting the operation situation of the power system based on the LSTM circulating neural network as set forth in claim 1, wherein the specific process of the step S1 is as follows:
s1-1, carrying out normalization processing on an initial data sample composed of power line operation data;
s1-2, performing sliding sampling on the initial data sample subjected to normalization processing to generate a final data sample;
s1-3, performing outlier processing on the final data sample.
3. The method for predicting the operation situation of an electric power system based on an LSTM recurrent neural network according to claim 2, wherein in the step S1-1, for any one electric power line, two ends of the electric power line are respectively marked as an I end and a J end; the initial data sample consists of operation data of the power line, and comprises 10 kinds of attribute data in total, wherein the sampling time interval is 1 hour.
4. The method for predicting the operation situation of an electric power system based on an LSTM recurrent neural network according to claim 2, wherein in said step S1-3, said outliers include blocks of zero and non-numerical values;
the outlier processing is as follows: the final data samples with outliers are screened out.
5. The method for predicting the operation situation of an electric power system based on an LSTM recurrent neural network according to claim 4, wherein in said step S2-1, the time window depth used by the operation state recognition module and the operation data trend prediction module is different, and the sliding window of the operation data trend prediction module is far deeper than the operation state recognition module.
6. The method for predicting the operation situation of an electric power system based on an LSTM recurrent neural network according to claim 5, wherein in step S3, the operation state recognition module performs training, verification and test using all 10 kinds of attribute data;
the operation data trend prediction module uses 4 kinds of attribute data including I-terminal active power, I-terminal reactive power, J-terminal active power and J-terminal reactive power to train, verify and test.
7. The method for predicting the operation situation of the power system based on the LSTM recurrent neural network as claimed in claim 6, wherein the specific process of step S3 is as follows:
s3-1, randomly scrambling the data samples processed in the S1 and dividing the data samples into a training set, a verification set and a test set according to the proportion;
s3-2, an operation state identification module adjusts the depth of the sliding time window and the super parameters of the network structure according to the identification accuracy of the test set;
s3-3, the operation data trend prediction module adjusts the sliding time window depth and the network structure super-parameters according to the error size of the predicted value;
s3-4, the running state identification module and the running data trend prediction module use respective data set samples for training and verification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311057728.3A CN117036103A (en) | 2023-08-22 | 2023-08-22 | Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311057728.3A CN117036103A (en) | 2023-08-22 | 2023-08-22 | Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117036103A true CN117036103A (en) | 2023-11-10 |
Family
ID=88642900
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311057728.3A Pending CN117036103A (en) | 2023-08-22 | 2023-08-22 | Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117036103A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117313018A (en) * | 2023-11-29 | 2023-12-29 | 国网浙江省电力有限公司 | Power transmission line abnormal state identification method and system |
-
2023
- 2023-08-22 CN CN202311057728.3A patent/CN117036103A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117313018A (en) * | 2023-11-29 | 2023-12-29 | 国网浙江省电力有限公司 | Power transmission line abnormal state identification method and system |
| CN117313018B (en) * | 2023-11-29 | 2024-01-30 | 国网浙江省电力有限公司 | Power transmission line abnormal state identification method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Capozzoli et al. | Fault detection analysis using data mining techniques for a cluster of smart office buildings | |
| CN114559819B (en) | Electric automobile battery safety early warning method based on signal processing | |
| Liu et al. | Sequence fault diagnosis for PEMFC water management subsystem using deep learning with t-SNE | |
| Zhang et al. | A novel fault diagnosis method based on stacked LSTM | |
| CN117036103A (en) | Electric power system operation situation prediction method based on LSTM (least squares) circulating neural network | |
| CN111273125A (en) | RST-CNN-based power cable channel fault diagnosis method | |
| CN107895212A (en) | Lead-acid battery life-span prediction method based on sliding window and various visual angles Fusion Features | |
| Sun et al. | A novel fault prediction method based on convolutional neural network and long short-term memory with correlation coefficient for lithium-ion battery | |
| CN113011248A (en) | Nuclear power plant fault diagnosis method based on LSTM | |
| Kim et al. | Outlier mining-based fault diagnosis for multiceli lithium-ion batteries using a low-priced microcontroller | |
| CN109615273A (en) | A kind of electric car electrically-charging equipment method for evaluating state and system | |
| CN117169761A (en) | Battery state evaluation method, apparatus, device, storage medium, and program product | |
| Li et al. | Battery safety risk assessment in real-world electric vehicles based on abnormal internal resistance using proposed robust estimation method and hybrid neural networks | |
| CN115146718A (en) | Depth representation-based wind turbine generator anomaly detection method | |
| CN105183659A (en) | Software system behavior anomaly detection method based on multi-level mode predication | |
| CN110956281A (en) | Power equipment abnormity detection alarm system based on Log analysis | |
| CN114252716A (en) | Power grid fault diagnosis method and device and storage medium | |
| CN208488118U (en) | A kind of intelligent transformer Integrated Fault Diagnosis System | |
| CN115718260A (en) | Battery fault prediction method and system based on combination of CNN-LSTM and correlation coefficient | |
| CN117805616A (en) | Abnormal cell identification method for lithium ion energy storage system | |
| CN110441725A (en) | Intelligent electric meter method for diagnosing faults and device based on Adaptive Neural-fuzzy Inference | |
| CN113949163A (en) | Anomaly monitoring system applying big data to smart power grid | |
| CN118011223B (en) | Fault diagnosis method and system for electric power energy storage system | |
| Zhao et al. | Reliable Life Prediction and Evaluation Analysis of Lithium-ion Battery Based on Long-Short Term Memory Model | |
| CN110837932A (en) | Thermal power prediction method of solar heat collection system based on DBN-GA model |
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 |