CN118152741A - A temperature situation awareness auxiliary control method for substation equipment - Google Patents
A temperature situation awareness auxiliary control method for substation equipment Download PDFInfo
- Publication number
- CN118152741A CN118152741A CN202410038804.4A CN202410038804A CN118152741A CN 118152741 A CN118152741 A CN 118152741A CN 202410038804 A CN202410038804 A CN 202410038804A CN 118152741 A CN118152741 A CN 118152741A
- Authority
- CN
- China
- Prior art keywords
- temperature
- data
- equipment
- model
- prediction
- 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.)
- Withdrawn
Links
Classifications
-
- H02J13/12—
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/0499—Feedforward networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Tourism & Hospitality (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明提供一种变电设备温度态势感知辅助控制方法，解决目前变电设备温度态势感知预测结果准确性与可信度低、未充分利用处理数据的多样性和异构性导致温度态势感知及决策可靠性低等问题。包括：对变电设备进行数据监测，采集变电设备关键点温度；筛选分类监测数据，弥补缺失温度数据，建立数据库；基于ARIMA与SVM的组合模型构建温度预测模型；分析变电设备温度集与故障集的关联水平，生成运行辅助决策方式。数据入库前剔除异常值补充缺失值，提升温度数据库完整性；结合ARIMA与SVM的温度态势预测模型，有效预判变电设备运行趋势；应用灰色关联分析法厘清设备温度集与历史故障集间关联性，奠定设备运行辅助决策基础，具有准确性高、可靠性强等优点。
The present invention provides a temperature situation perception auxiliary control method for substation equipment, which solves the problems of low accuracy and credibility of the current prediction results of temperature situation perception of substation equipment, low reliability of temperature situation perception and decision-making due to insufficient utilization of the diversity and heterogeneity of processed data. It includes: monitoring the data of substation equipment, collecting the temperature of key points of substation equipment; screening and classifying monitoring data, making up for missing temperature data, and establishing a database; building a temperature prediction model based on the combined model of ARIMA and SVM; analyzing the correlation level between the temperature set and the fault set of substation equipment, and generating an operation auxiliary decision-making method. Before data is stored in the database, outliers are eliminated and missing values are supplemented to improve the integrity of the temperature database; the temperature situation prediction model of ARIMA and SVM is combined to effectively predict the operation trend of substation equipment; the gray correlation analysis method is used to clarify the correlation between the equipment temperature set and the historical fault set, laying the foundation for equipment operation auxiliary decision-making, and has the advantages of high accuracy and strong reliability.
Description
Technical Field
The invention relates to the technical field of operation and maintenance of power transformation equipment, in particular to a power transformation equipment temperature situation awareness auxiliary control method.
Background
The transformer substation is an important link of the power system, and the urgent need of ensuring the safe and stable operation of the transformer equipment and improving the intelligent management level of the transformer equipment is present. In the aspect of abnormal heating faults or early warning in the operation of substation equipment, intensive research is needed from a plurality of aspects of operation, management and control. By accurately judging the future operation state of the power transformation equipment, a reference can be provided for formulating a reasonable operation mode, and the safe and reliable operation of the power system can be effectively supported. At present, the existing related researches at home and abroad mainly focus on three aspects of load or temperature prediction, a power equipment fault early warning method and auxiliary control of equipment operation modes.
The load or temperature prediction mainly researches the connection between the heating fault of the electrical equipment of the transformer substation and the equipment temperature, provides a reference for the operation management of the accurate equipment, and realizes the early warning of the temperature fault by carrying out data mining on a temperature data set and predicting the future temperature change rule of the equipment by adopting a related prediction method based on the historical temperature data change rule. The common prediction method mainly comprises methods of time sequence, chaos fuzzy theory, deep learning and the like. The power equipment fault early warning method mainly predicts relevant influence factors of a target through supervision learning analysis, adopts an intelligent prediction model, is based on data mining and big data analysis technology, and achieves future equipment operation risk prediction through mining and analysis of fault prediction and diagnosis data. The fault early warning mainly depends on a fault diagnosis result, and if the fault diagnosis method is selected improperly, false alarm or missing alarm of an early warning system can be caused, so that the safety and the stability of the equipment are negatively affected. Currently, fault diagnosis methods include a surface temperature judgment method, a similar comparison judgment method, and a relative temperature difference judgment method. Regarding the auxiliary control aspect of the equipment operation mode, the method is mainly based on real-time monitoring data of the substation equipment operation information, deep mining analysis is carried out on the collected data, and health state assessment, life prediction and maintenance time prediction are further carried out on the equipment so as to make an accurate maintenance plan.
Analysis of related researches on the intelligent management level of the existing power transformation equipment mainly comprises the following problems: in the original temperature data processing process, the influence of different data types and the correlation conduction of data outliers or noise among different data is not fully considered, and meanwhile, the accuracy and the reliability of a prediction result are reduced due to the fact that historical data are missing. In the aspect of fault diagnosis method selection, the surface temperature judgment method is used for effectively identifying the temperature change condition inside the equipment, the diagnosis result is easily influenced by direct sunlight or coating materials, the similar comparison judgment method compares the temperature values of the corresponding parts of the similar equipment, but the reference equipment is easily influenced by the environmental temperature, so that misjudgment or missed judgment of the diagnosis result occurs; the relative temperature difference judging method can diagnose the temperature difference between different devices or areas, but the diagnosis result is greatly influenced by the environmental temperature, and if the environmental temperature is unstable, the judging result has errors. Since the underlying data originates from different types of power transformation devices or information networks and covers different time periods, further intensive research is still needed in terms of the diversity and isomerism of the processed data, as well as the implementation of orderly integration.
The Chinese patent literature discloses an operation and maintenance auxiliary control method for power transformation equipment, wherein the publication number of the auxiliary control method is CN115345379A, and the control method defines the health index of the power transformation equipment by dividing the health state of the power transformation equipment into a plurality of state grades; calculating the health coefficient of the equipment by adopting a long-short-term memory neural network LSTM; calculating the health coefficient of the equipment by adopting a Bert-BiLSTM neural network; calculating a health index critical value of the power transformation equipment, and formulating a corresponding operation and maintenance strategy for each health state grade of the power transformation equipment; and finally, respectively taking the structured data and unstructured text data of the power transformation equipment to be subjected to operation and maintenance auxiliary control as inputs of the LSTM neural network and the Bert-BiLSTM neural network model which are completed by training, calculating the health index of the power transformation equipment, and providing auxiliary control for the operation and maintenance of the power transformation equipment. The control method provided by the invention does not carry out high-efficiency sensing on the temperature situation of the power transformation equipment, the effectiveness and the integrity of the data are ensured by not inputting abnormal data through the processing model, the data processing method is improved, and the accuracy of the high-efficiency sensing on the running state of the power transformation equipment is still to be improved.
Disclosure of Invention
The invention aims to solve the problems that the accuracy and the reliability of the current temperature situation awareness prediction result of the power transformation equipment are low, the diversity and the isomerism of the processed data are not fully utilized, the temperature situation awareness and the control reliability of the power transformation equipment are low, and the like.
The technical problems are solved by the following technical scheme: a transformer equipment temperature situation awareness auxiliary control method comprises the following steps:
s1, monitoring data of power transformation equipment and acquiring temperature of key points of the power transformation equipment;
s2, screening and classifying the monitoring data, and building a substation equipment monitoring database by using an associated transfer model to make up for the historical temperature data loss of the key points of the original substation equipment;
s3, constructing a transformation equipment temperature prediction model based on a combined model of ARIMA and SVM;
And S4, analyzing the association level between the temperature set of the power transformation equipment and the fault set of the power transformation equipment according to a gray association analysis method, and generating an operation auxiliary control mode.
The power transformation equipment temperature situation perception auxiliary control method is mainly developed from 4 aspects of a perception layer, an understanding layer, a prediction layer and an auxiliary control layer. And at the perception layer and the understanding layer, researching the relevance among multiple types of temperature data by using a K neighbor classification algorithm, and under the condition that the historical temperature data is missing, constructing a device historical data transfer model by using a BP neural network so as to improve the integrity and reliability of a database. Meanwhile, in order to make up for the defect of the autoregressive integral moving average model in the aspect of processing nonlinear data and noise, a temperature prediction model combining ARIMA and a support vector machine SVM is designed. And finally, researching the correlation level between the relative temperature difference and the temperature rise of the equipment and the equipment fault set by using a gray correlation analysis method, designing an equipment operation auxiliary control scheme, and providing a control basis for adjusting the operation mode of the power transformation equipment and making an overhaul plan. Abnormal values are removed and missing values are supplemented by using a KNN neighbor and BP neural network method, so that the integrity of a temperature database is improved, and accurate input data are provided for temperature prediction. The device temperature situation prediction model based on the combination of ARIMA and SVM provides support for accurate analysis of future temperature change trend and intelligent auxiliary control, effectively pre-judges the operation trend of the power transformation device, and provides basis for formulating a safe operation strategy of the device. The gray correlation analysis method is utilized to clear the correlation relation between the equipment temperature set and the historical fault set, so that a theoretical basis is laid for auxiliary control of equipment operation, and a reference basis is provided for equipment operation management.
Preferably, in step S3, constructing the transformation device temperature prediction model based on the combined model of ARIMA and SVM includes:
S301, modeling a linear part by using an ARIMA model, performing differential operation on a temperature data time sequence, converting the time sequence into a stable time sequence, analyzing an autocorrelation function and a partial autocorrelation function of the stable time sequence, and determining autoregressive and moving average orders of the ARIMA model;
s302, fitting time sequence data by using an ARIMA model, generating a prediction result of a linear autocorrelation component, and acquiring a residual sequence of the ARIMA model according to the linear prediction result;
S303, reconstructing a residual sequence to obtain an SVM sample set, and adopting an SVM model to process nonlinearity and noise and predict residual;
S304, combining the prediction result of the linear autocorrelation component with the residual prediction result to obtain a model prediction result, and checking a prediction model by using the test set data.
The ARIMA model is utilized to realize the mining of the association relation of different data with time change, and the data in the actual production process is generally influenced by noise or other interference, so that the calculated amount of the model is increased. In order to solve the defects of the ARIMA method in the aspect of processing nonlinear and noise data, a complementary mode of combining the ARIMA and the SVM is adopted, namely, the trend and the periodicity of the temperature change along with time are analyzed by adopting an ARIMA model, the noise and nonlinear problems are effectively processed by utilizing the SVM, the operation trend of the power transformation equipment is effectively prejudged, and the scientificity and the reliability of the safety operation strategy of the formulated equipment are improved.
Preferably, in step S305, verifying the prediction model using the test set data includes: when the prediction result is not ideal, repeatedly adjusting model parameters, and retraining a prediction model, wherein the model parameters comprise the ARIMA model order and parameters of an SVM kernel function; when the prediction result is ideal, the prediction performance of the model is evaluated by using the root mean square error and the average absolute error. When the prediction model is checked, the model parameters are adjusted for a plurality of times, so that the less ideal prediction model is retrained, and the model performance is improved.
Preferably, in step S4, analyzing the correlation level between the substation equipment temperature set and the substation equipment fault set according to the gray correlation analysis method includes: constructing a device temperature set by using the relative temperature difference and the temperature rise of the power transformation device; constructing a historical temperature fault set through fault data in the historical data, wherein the historical temperature fault set comprises temperature data in the equipment fault state; and calculating gray correlation degree between the temperature data set and the fault set, and grading the running state of the equipment according to the magnitude of the correlation degree value. And the gray correlation degree analysis reflects the correlation degree between the data, and when the gray correlation value is larger, the correlation degree of the data is higher, so that the relative temperature difference and the temperature rise of the equipment are closer to the running state when faults occur. Accordingly, the operation state of the device can be classified according to the magnitude of the gray correlation value.
Preferably, calculating the gray correlation between the temperature dataset and the fault set comprises: analyzing the temperature of the power transformation equipment in an early warning way, calculating the relative temperature difference delta and the temperature rise delta T of the equipment, and constructing a temperature set of the equipment according to the relative temperature difference and the temperature rise; constructing a historical temperature fault set of different equipment according to the historical temperature data, wherein the historical temperature fault set comprises relative temperature difference and temperature rise corresponding to equipment faults; carrying out normalization processing on the temperature set and the fault set, and calculating the association degree between the temperature set and the fault set by using the gray association degree, wherein the gray association degree calculation formula is as follows:
wherein: r is the gray correlation degree of the relative temperature difference and temperature rise of the equipment and the fault set; ρ is the resolution factor; delta c is the corresponding relative temperature difference when the equipment fails; delta h is the relative temperature difference of the equipment during early warning analysis.
Preferably, in step S2, compensating for the historical temperature data loss of the key points of the original power transformation device includes: training a transmission model of the BP neural network; establishing an association transfer model between the change of the equipment operation historical data and the historical temperature data; and storing the repair data output by the model into a monitoring database. The BP neural network is a multi-layer feedforward network consisting of an input layer, an hidden layer and an output layer, and a correlation transfer model between the change of operation historical data such as load, humidity, light intensity and wind speed and the like and the historical temperature data is established, so that the defect and the deficiency of the historical temperature data of key points of the original power transformation equipment are overcome, and the BP neural network is mainly realized through a BP neural network method.
Preferably, training the transfer model of the BP neural network includes: dividing data into a training set and a testing set according to a preset proportion; inputting training set data into a model for training, comparing training results with historical data in a test set, and calculating a prediction error of the model; during the model training process, back propagation is performed and the whole process is circularly performed until the error falls below the allowable error or the set training times are reached. The error value is reversely transferred to each layer along the network through the back propagation BP neural network so as to further train the network and reduce the error.
Preferably, in step S2, the screening and classifying the monitored data includes: classifying data by adopting a KNN algorithm, and classifying the data into different categories by utilizing an Euclidean distance method; assigning data to be put in storage to the class with the most neighbors; before the data is stored in the database, the respective threshold value of each group of category data is determined according to the distribution condition of the known data points, and abnormal values are removed. Before data is stored in a database, abnormal value identification and elimination operation are executed, so that the accuracy of the database data is improved.
Preferably, in step S1, the data monitoring of the power transformation device includes data monitoring of a voltage-induced thermal device including a voltage transformer and a coupling capacitor and a current-induced thermal device including a junction contact portion of an isolating switch and a circuit breaker, the monitored data being represented in a time-series form. The design of the transformer equipment temperature monitoring database mainly carries out big data analysis and excavation on basic information collected by the temperature measuring robot so as to share collected information resources and construct a platform temperature monitoring database.
The beneficial effects of the invention are as follows: the invention builds a research framework of a temperature situation sensing and auxiliary control system of the power transformation equipment, and discusses feasibility of the power transformation equipment on four layers of a sensing layer, an understanding layer, a prediction layer and auxiliary control. When the warehouse-in data is cleaned, abnormal values are removed and missing values are supplemented by utilizing a KNN neighbor and BP neural network method, so that the integrity of a temperature database is improved, and accurate input data is provided for temperature prediction; the device temperature situation prediction model based on the combination of ARIMA and SVM can provide support for accurate analysis of future temperature change trend and intelligent auxiliary control, effectively pre-judge the operation trend of the power transformation device and provide basis for formulating a safe operation strategy of the device; meanwhile, the association relation between the equipment temperature set and the historical fault set can be clarified by using a gray association analysis method, so that a theoretical basis is laid for auxiliary control of equipment operation, and a reference basis is provided for equipment operation management.
Drawings
FIG. 1 is a schematic diagram of a layered framework of the method of the present invention.
FIG. 2 is a flow chart of a method of constructing a temperature prediction model in accordance with the present invention in combination with ARIMA and SVM models.
Detailed Description
Embodiment one: the embodiment provides a transformer equipment temperature situation awareness auxiliary control method, which comprises the following steps:
s1, monitoring data of power transformation equipment and acquiring temperature of key points of the power transformation equipment;
s2, screening and classifying the monitoring data, and building a substation equipment monitoring database by using an associated transfer model to make up for the historical temperature data loss of the key points of the original substation equipment;
s3, constructing a transformation equipment temperature prediction model based on a combined model of ARIMA and SVM;
And S4, analyzing the association level between the temperature set of the power transformation equipment and the fault set of the power transformation equipment according to a gray association analysis method, and generating an operation auxiliary control mode.
In S1, the substation equipment information monitoring data is mainly classified into the following two types: one is voltage-heating type equipment such as voltage transformers, insulators, coupling capacitors, cable heads and the like; and the second is a current heating type device which comprises a disconnecting switch, a circuit breaker and other joint contact points. The collected temperature monitoring data is represented in time series form, such as: y t＝{ti,Ti }, wherein t i is a device i timestamp, representing a specific time corresponding to temperature acquisition; t i represents the device temperature value measured by device i at time T i.
In S2, for the historical data and the real-time temperature data, the KNN algorithm is first used to classify the data, and the euclidean distance method is used to classify the data into different categories [15]. Two points in the data feature spaceThe distance between them is expressed as:
Wherein: the temperature data to be put in storage; /(I) Is the stored temperature data.
The data x k to be binned is assigned to the class with the most nearest neighbor. Outlier identification and culling operations must be performed before the data is stored in the database. According to the distribution condition of known data points, each group of class data needs to determine different thresholds, and a basis is provided for eliminating abnormal values which deviate from the class center farther.
And establishing an associated transfer model between the change of the equipment operation historical data and the historical temperature data, wherein the historical data comprises load, humidity, light intensity, wind speed and the like. The method is characterized in that the deficiency and the deficiency of historical temperature data of key points of original power transformation equipment are mainly realized through a BP neural network method, wherein the BP neural network comprises an input layer, an hidden layer and an output layer, and a multi-layer feedforward network is formed by the following calculation formulas from the input layer to the hidden layer and from the hidden layer to the output layer:
Wherein: x i is input data of the neural network; w 1i,W2i are respectively the connection weights; n is the number of historical data samples; b 1 and b 2 are thresholds; f (S l) is the output result of the hidden layer; s m is the output layer output result.
In order to improve the accuracy of model prediction and consider the data size of the database, the data can be divided into a training set and a testing set according to the ratio of 3:1. The training set data is used as input, the historical data in the test set is used as expected output, the training set data is input into the model for training, the result is compared with the expected output, and the prediction error of the model is calculated to verify the prediction accuracy of the model and provide basis for the adjustment of model parameters. In the model training process, forward propagation means that original data is firstly transferred from an input layer to a hidden layer, then transferred to an output layer through the hidden layer, and finally the result is output to the outside. The temperature data output by the output layer is to be compared with expected data, if a larger prediction error exists, back propagation is needed, and the purpose of the back propagation is to further train the network by reversely transmitting the error value to each layer along the network, so that the error is reduced. The whole process is circularly carried out until the error is reduced below the allowable error or the set training times are reached; after the transmission model of the BP neural network is trained, the temperature data of the key points with defects in the database can be repaired by using the model, and the repair data are stored in the database.
In S3, the ARIMA model is utilized to mine the association relationship of different data with time, and in the actual production process, the data is generally affected by noise or other interference, so that the calculation amount of the model increases. Combining ARIMA with SVM can solve the shortcomings of ARIMA method in processing non-linearity and noise data, adopting ARIMA model to analyze trend and periodicity of temperature change along with time, using SVM to effectively process noise and non-linearity,
The ARIMA model consists of three parts, namely autoregressive, differential and moving average, wherein AR represents an autoregressive part, I represents a differential operation part and MA represents a moving average part. The model is denoted ARIMA (p, d, q), where p, d, q represent the autoregressive order, the differential order, and the moving average order [20-21], respectively. Firstly, the temperature data time sequence is processed into a stable sequence by d-level difference processing,
Yt (d)＝(1-L)dYt
Wherein: y t is the original time sequence; y t (d) is a time series after d-order differential processing. L is a hysteresis operator, denoted as L jYt＝Yt-j, and d is the differential order.
The autoregressive portion represents a linear relationship between the current time point temperature value and the historical time point temperature value, and the mathematical model thereof can be expressed as:
Wherein: is an autoregressive coefficient; e t is a white noise error term.
The moving average portion represents a linear combination of the current point-in-time temperature value and the first few white noise error terms, and its mathematical model can be expressed as:
Yt＝θ1et-1+θ2et-2+...+θqet-q+et
Wherein: θ 1,θ2...θq is a running average coefficient.
The autoregressive moving average model ARMA (p, q) can be obtained by adding the autoregressive part and the moving average part, and the specific expression is as follows:
Where C is a constant, ARIMA (p, d, q) is compared to ARMA (p, q), and the ARIMA model has more differential operation parts than the ARMA model. The temperature time sequence Y t in the ARIMA formula is subjected to d-order differential processing, so that trend or periodic components in the sequence are reduced, and a time sequence dataset can be converted into a stable differential sequence, so that the model is easier to converge, and the model prediction precision is improved.
Regarding the SVM model, firstly, a kernel function of the SVM model is adopted to map data to a high-dimensional feature space, and inner product operation is carried out in the feature space so as to analyze nonlinear relation in the data. The mathematical expression is as follows:
f(x)＝ωTΦ(x)+b
wherein: omega is the normal vector; b is a bias term; Φ (x) is a map of the input temperature data x.
The optimization problem can be obtained by using the maximum spacing method:
s.t.Yi＝ωTΦ(x)+b+ξi
wherein: c is a penalty coefficient; xi i is a non-negative relaxation factor; ω is the normal vector.
Finally, the SVM regression estimation function expression is:
Wherein: m is the number of training samples; a i is a weight coefficient obtained in the training process; x, x i is the input feature of the training sample, and K (x, x i) is the kernel function.
In order to analyze a complex nonlinear relation in temperature data, a radial basis function RBF kernel is adopted for modeling prediction, and the expression is:
k(x,xi)＝exp(-γ||x-xi||2)
Wherein: x, x i are input samples; gamma is a parameter of the RBF core.
In S4, the gray correlation method may be used to analyze the similarity between different data, where a larger value indicates a higher correlation of data. Firstly, constructing a device temperature set by utilizing relative temperature difference and temperature rise of the device, then constructing a historical temperature fault set by fault data in the historical data, wherein the historical temperature fault set comprises temperature data in a device fault state, and finally, calculating gray correlation between the temperature data set and the fault set, and grading the running state of the device by the magnitude of the correlation value.
The temperature early warning analysis of the power transformation equipment is as follows:
The relative temperature difference is used for calculating the proportional relation between the temperature difference and the temperature rise between the two temperature measuring points, when the relative temperature difference is calculated, the environment of the two temperature measuring points is required to be ensured to be similar to the basic condition of equipment, the equipment model, the installation site, the load information, the surface insulating material and the like, and the calculation method of the relative temperature difference is as follows:
Wherein: t 1、T2 is the temperature of the two measuring points respectively; t 0 is the ambient temperature; delta T 1、ΔT2 is the equipment temperature rise respectively; Δt is the temperature rise relative to the environment.
The calculation mode of the equipment temperature rise is that the measured equipment temperature T a at the time a is subtracted by the ambient temperature T a0 at the time a, and the average ambient air temperature in 24 hours is required not to exceed 35 ℃:
ΔT＝Ta-Ta0
The relative temperature difference delta and the temperature rise delta T of the equipment can be calculated through the method, the temperature set Z h＝{δh,ΔTh of the equipment is constructed according to the relative temperature difference and the temperature rise, meanwhile, the historical temperature fault set Z C＝{δc,ΔTc of different equipment is constructed through historical temperature data, and the historical temperature fault set contains the relative temperature difference and the temperature rise corresponding to equipment faults.
And carrying out normalization processing on the temperature set and the fault set, and then calculating the association degree between the temperature set and the fault set by using the gray association degree, wherein the gray association degree calculation method comprises the following steps:
Wherein: r is the gray correlation degree of the relative temperature difference and temperature rise of the equipment and the fault set; ρ is the resolution factor.
By utilizing gray correlation analysis, the correlation degree between data can be reflected, and when the gray correlation value is larger, the correlation degree of the data is higher, and the relative temperature difference and the temperature rise of the equipment are closer to the running state when faults occur. Accordingly, the operation state of the device may be classified according to the magnitude of the gray correlation value, and corresponding measures may be taken as shown in table 1 below.
TABLE 1 hierarchical tables of device states
Embodiment two: the embodiment provides a practical method for combining ARIMA and SVM to construct a temperature prediction model, which adopts the following technical means: as shown in fig. 2, the temperature time sequence Y t can be regarded as a combination of two parts of a linear autocorrelation part L t and a nonlinear residual N t, namely Y t＝Lt+Nt, and the specific steps are as follows:
Modeling the linear part by using an ARIMA model, performing differential operation on the temperature data time sequence, and converting the time sequence into a stable time sequence; the plateau time series is analyzed for autocorrelation and partial autocorrelation functions to determine the autoregressive and moving average orders of the ARIMA model.
Fitting the time series data using the ARIMA model will generate a prediction result L t' of the linear autocorrelation component. And the residual sequence of the ARIMA model is obtained by utilizing the linear prediction result, the specific calculation method is shown in the formula 12, and the residual sequence N t' reflects the nonlinear relation in the sequence.
Nt'＝Yt-Lt'
Reconstructing the N t 'sequence obtained in the last step to obtain an SVM sample set, wherein N t' comprises the nonlinear relation of data and noise, processing the nonlinearity and the noise by adopting an SVM model, and predicting residual errors to obtain a prediction result N t ".
And combining L t 'obtained by linear prediction with N t' obtained by nonlinear aggregation to obtain a prediction result Y t'＝Lt'+Nt ".
And checking the prediction model, checking the model by using the test set data, and evaluating the prediction performance of the model by using root mean square error and average absolute error. If the prediction result is not ideal, the model parameters including ARIMA model order, SVM kernel parameters and the like are tried to be adjusted, and the prediction model is retrained to improve the model performance.
The specific embodiments described herein are offered by way of illustration only. Those skilled in the art may make various modifications or additions to the described embodiments or substitutions thereof without departing from the spirit of the application or exceeding the scope of the application as defined in the accompanying claims.
Although the present application uses more terms such as temperature fault set, gray correlation value, etc., the possibility of using other terms is not excluded. These terms are used merely for convenience in describing and explaining the nature of the application; they are to be interpreted as any additional limitation that is not inconsistent with the spirit of the present application.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410038804.4A CN118152741A (en) | 2024-01-10 | 2024-01-10 | A temperature situation awareness auxiliary control method for substation equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410038804.4A CN118152741A (en) | 2024-01-10 | 2024-01-10 | A temperature situation awareness auxiliary control method for substation equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118152741A true CN118152741A (en) | 2024-06-07 |
Family
ID=91284181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410038804.4A Withdrawn CN118152741A (en) | 2024-01-10 | 2024-01-10 | A temperature situation awareness auxiliary control method for substation equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118152741A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119067265A (en) * | 2024-11-04 | 2024-12-03 | 无锡广盈集团有限公司 | A predictive maintenance method and system for electric power facilities |
| CN119354280A (en) * | 2024-12-25 | 2025-01-24 | 南京市计量监督检测院 | Large-scale spatial temperature and humidity assessment method based on multi-information fusion |
-
2024
- 2024-01-10 CN CN202410038804.4A patent/CN118152741A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119067265A (en) * | 2024-11-04 | 2024-12-03 | 无锡广盈集团有限公司 | A predictive maintenance method and system for electric power facilities |
| CN119354280A (en) * | 2024-12-25 | 2025-01-24 | 南京市计量监督检测院 | Large-scale spatial temperature and humidity assessment method based on multi-information fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Udo et al. | Data-driven predictive maintenance of wind turbine based on SCADA data | |
| CN115833400B (en) | Monitoring and early warning method and system for power equipment of transformer substation | |
| CN118739184B (en) | A method for automatic fault detection and repair of self-healing intelligent power lines | |
| CN105891629B (en) | A kind of discrimination method of transformer equipment failure | |
| CN118157132B (en) | Data mining method and device for voltage monitoring system based on neural network | |
| CN113156917A (en) | Power grid equipment fault diagnosis method and system based on artificial intelligence | |
| CN119167805B (en) | Power failure calculation prediction method and system based on digital twin | |
| CN106017551A (en) | Intelligent transmission line integrated monitoring analysis and early warning method | |
| WO2023065580A1 (en) | Fault diagnosis method and apparatus for gearbox of wind turbine generator set | |
| CN118152741A (en) | A temperature situation awareness auxiliary control method for substation equipment | |
| CN118861662B (en) | A method for fault detection of oil-immersed transformers based on gas on the oil surface | |
| CN114519923B (en) | Intelligent diagnosis and early warning method and system for power plant | |
| CN116050888A (en) | A method for evaluating the health status of sensors in intelligent high-voltage switchgear | |
| CN108627720A (en) | A kind of power equipment state monitoring method based on bayesian algorithm | |
| CN110009208A (en) | A kind of on-pole switch complete set of equipments health state evaluation method and device based on integrated intelligent algorithm | |
| CN120721162B (en) | Operation, maintenance, commissioning and monitoring system suitable for electric fields | |
| CN114252103B (en) | Fusion power station operation fault prediction method | |
| CN117639228A (en) | Distribution network operating status prediction method and system based on digital twins | |
| CN113466597A (en) | Intelligent detection method for equipment state of rail transit power supply system | |
| CN118311352A (en) | Bus duct fault diagnosis method and system for photovoltaic energy storage system | |
| CN115511263B (en) | Switchgear safety risk prediction method based on CEEMDAN and BiLSTM model | |
| CN119916130A (en) | Real-time analysis and precise positioning of distribution network faults and their treatment methods and systems | |
| CN106292631A (en) | A kind of PWM rectifier fault diagnosis system based on neutral net | |
| CN119784302A (en) | A prediction and warning system based on the galloping characteristics of transmission lines under complex meteorological conditions | |
| CN116595319A (en) | Prediction method and system applied to rail transit motor health status assessment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20240607 |