CN112182956A - SF based on BP neural network6Method for predicting pressure of online monitoring device - Google Patents
SF based on BP neural network6Method for predicting pressure of online monitoring device Download PDFInfo
- Publication number
- CN112182956A CN112182956A CN202010944016.3A CN202010944016A CN112182956A CN 112182956 A CN112182956 A CN 112182956A CN 202010944016 A CN202010944016 A CN 202010944016A CN 112182956 A CN112182956 A CN 112182956A
- Authority
- CN
- China
- Prior art keywords
- pressure
- neural network
- monitoring device
- predicting
- data
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01L—MEASURING FORCE, STRESS, TORQUE, WORK, MECHANICAL POWER, MECHANICAL EFFICIENCY, OR FLUID PRESSURE
- G01L19/00—Details of, or accessories for, apparatus for measuring steady or quasi-steady pressure of a fluent medium insofar as such details or accessories are not special to particular types of pressure gauges
- G01L19/08—Means for indicating or recording, e.g. for remote indication
- G01L19/12—Alarms or signals
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to SF based on BP neural network6The pressure prediction method comprises analyzing environmental temperature and its variation, relative humidity, wind speed, weather type and conductor current for SF6Influence of pressure change, establishing BP neural network prediction model, and applying to SF6And predicting the pressure value of the online monitoring device, providing data reference for normal operation of monitoring equipment, and providing a judgment strategy for analyzing the GIS equipment pressure reduction reason. Selecting typical SF of GIS equipment of certain extra-high voltage transformer substation6On-line monitoring deviceMATLAB simulation is carried out on gas pressure, corresponding natural environment and conductor current data, and the result shows that SF6The accuracy of the pressure prediction result can reach more than 98.5%, and the correctness of theoretical analysis is verified.
Description
Technical Field
The invention belongs to the technical field of electric power operation and maintenance equipment, and particularly relates to SF (sulfur hexafluoride) based on a BP (back propagation) neural network6A method for predicting the pressure of an online monitoring device.
Background
SF6The gas is widely used in SF due to its good electrical insulation and excellent arc extinguishing ability6Circuit breakers, enclosed switchgear (GIS) and the like. During operation of the plant, SF6Gas density is one of the important factors determining the arc extinguishing and insulating capability of the circuit breaker, and can pass through SF6Density relay reading SF of corresponding equipment6Pressure, or by addition of SF6The on-line monitoring device realizes the SF control6Real-time tracking and recording of pressure.
By external natural environment and internal conductorCurrent, SF shown in both devices6Non-linear fluctuations of the pressure data, if displayed SF6When the pressure is lower than the preset value, the system can send out an alarm or a locking signal by mistake, and the normal operation of the equipment is influenced. Therefore, it is necessary to provide the device SF6The pressure change rule is researched and predicted, so that the SF of the equipment at a certain moment is analyzed and judged6Whether the pressure change is a normal phenomenon or not improves the reliability of equipment operation.
Zhang Cheng et al for SF together6Porcelain column type circuit breaker SF6The low air pressure alarm event is researched to point out that SF is required to be ensured6The density relay is consistent with or close to the environment of the main air chamber, and the compensation function of the density relay can be correctly realized; benefit-discriminating person mistakenly sends SF by aiming at 220kV circuit breaker6The analysis of the reason of the low gas pressure alarm signal provides that the density relay can sense the ambient temperature of the breaker body by adopting a temperature sensing bulb method, thereby avoiding the mistaken sending of the pressure locking signal. Wangxiang bin studied the severe cold area in winter due to SF6The problem of circuit breaker locking caused by liquefaction is solved by adopting a method for temporarily heating the circuit breaker; plum sea wave summarizes various types of SF6The density controller has the advantages that the structure characteristics of the density controller qualitatively analyze the influence of factors such as temperature compensation mode, altitude, temperature rise of electrical equipment, meter oil leakage and the like on accurate gas monitoring, and related cautions are provided by combining product design and operation maintenance. Chen Yuanming research on SF reduction6The compensation method for the measurement error of the online monitoring device is verified by using test data. The current research is mainly from SF6Qualitative judgment of pressure change, reduction of measurement error, lack of SF for actual operation6Quantitative analysis and prediction of the change in the pressure of the gas chamber.
Disclosure of Invention
The invention aims to provide SF based on BP neural network6An on-line pressure monitoring and predicting method for selecting a certain type SF of 1000kV GIS equipment6On-line monitoring device pressure is a research object, and factors such as external natural environment and internal current are analyzed for SF6Variation of pressureEstablishing BP neural network prediction model for SF6And (4) predicting the pressure, providing a discrimination strategy for analyzing the pressure reduction reason of the GIS equipment, and providing reference for monitoring the running state of the equipment and identifying the pressure reduction reason.
The invention adopts the following technical scheme:
SF based on BP neural network6The prediction method of the pressure of the online monitoring device comprises the following steps:
(1) data selection and processing: selecting pressure data acquired by an online monitoring device of a phase air chamber to be detected of an interval circuit breaker, internal current data of the interval circuit breaker at a corresponding moment and environment data at the corresponding moment;
(2) designing a BP neural network model;
(3) BP neural network training
(4) BP neural network pair SF6And predicting and analyzing the pressure of the online monitoring device.
In the step (1), various data are normalized by using a mapminmax function, and the value range of each variable is [ -1, 1 ].
In the step (1), the environmental data includes environmental temperature, relative humidity, weather type, wind speed, and temperature change rate.
In step (1), the sampling interval was 15 minutes.
In the step (2), the BP neural network model adopts a 3-layer topological structure BP neural network, 6 input layer nodes are respectively environment temperature, relative humidity, weather type, wind speed, temperature change rate and conductor current; the number of output layer nodes is 1.
In the step (2), the transfer function of the hidden layer is set to tansig, the transfer function of the output layer is set to logsig, and the training function is set to trainlm; the number of hidden layer nodes is 13.
And (3) performing simulation training by using a BP neural network toolbox in MATLAB.
In the step (4), an error threshold value P is set by utilizing pressure error data obtained by predicting a BP neural network model0The effectiveness of the measurement error is measured, and the actual state of the equipment is further processedAnd (6) analyzing the rows.
In the step (4), when the equipment generates an air pressure low alarm signal, the air chamber SF can be subjected to continuous observation on the difference relation between the prediction result P' and the actual pressure P6The pressure drop is identified as the cause.
In the step (4), the result P' obtained by the prediction of the BP neural network is compared with the actual pressure value P: if the difference between P' and P is substantially less than P within 2 consecutive hours0If the air chamber pressure data is normal, the air chamber pressure change is caused by external environment change, and air leakage fault of equipment does not occur; if the difference between P' and P is substantially larger than P within 2 hours of the reaction0And the P value shows a descending trend along with the time change, and the equipment can be judged to be air leakage.
The invention has the beneficial effects that: the invention mainly researches SF based on BP neural network6The pressure predicting method for the on-line monitoring device analyzes the factors such as the external natural environment, the current of the internal conductor and the like to SF6Influence of pressure change, establishing BP neural network prediction model, and applying to SF6And predicting the pressure value of the online monitoring device to provide data reference for normal operation of the monitoring equipment. Selecting typical SF of GIS equipment of certain extra-high voltage transformer substation6The gas pressure data of the on-line monitoring device and the corresponding natural environment and conductor current data are subjected to simulation analysis, and the result shows that: compared with the actual value, the prediction accuracy of the pressure prediction value can reach more than 98.5 percent, and the effectiveness of the prediction model is verified. On the basis, the method provides the analysis of GIS equipment SF6The judgment strategy of the pressure reduction reason provides powerful support for rapidly analyzing the pressure reduction reason of the equipment and improving the operation reliability of the equipment.
Drawings
FIG. 1 is a typical SF of GIS breaker on-line monitoring device6Pressure curve.
FIG. 2 shows SF6Analysis chart of influence factors of pressure change.
FIG. 3 shows the influence factors and SF6And monitoring a correlation result graph of the pressure values on line.
Fig. 4 is a topology structure diagram of the BP neural network.
FIG. 5 is a BP neural network prediction model.
Fig. 6 shows the BP network training results.
FIG. 7 is a graph of regression analysis.
FIG. 8 is a BP neural network overall prediction error graph.
FIG. 9 is a Q-Q diagram of a normal distribution.
Fig. 10 is a prediction error distribution histogram.
FIG. 11 is a BP neural network prediction graph.
Detailed Description
The technical solutions are described below clearly and completely with reference to the embodiments of the present invention and the accompanying drawings, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
1、SF6Density monitoring device
Existing SF6The density monitoring device mainly comprises SF6Density relay and SF6The two online monitoring devices respectively adopt different temperature compensation modes.
SF6The temperature compensation principle is as follows: SF in high voltage electrical equipment6The gas is sealed in a fixed container, has a certain density under the rated pressure at 20 ℃, and under the condition that the equipment is not sealed by gas, the gas pressure changes along with the temperature change, but the density of the gas is not changed all the time. In order to effectively monitor whether air leakage occurs or not, the measured real-time pressure is converted into the corresponding pressure at 20 ℃, and the converted pressure can be used as SF6The gas density, which is essentially the gas pressure at 20 ℃, represents the gas content in the fixed volume.
1.1SF6Density relay
Currently, SF is used in high voltage substations6Gas density relays are mostly of the mechanical pointer type.
Pointer SF6The density relay is realized by the expansion and contraction of a bimetallic strip in a dialTemperature compensation is realized. Because the dial plate volume is less, although density relay and equipment are in under same ambient temperature, its dial plate temperature change speed is far higher than the body, in addition, the influence of solarization radiation change to dial plate temperature also can not neglect. The pressure data of the device of the type needs manual reading, the data precision is low, the data is not easy to continuously collect, and the device can be used for long-term data recording and comparison and is not suitable for short-term prediction analysis.
1.2SF6Density on-line monitoring device
With the development of intelligent monitoring technology, SF with high data precision and real-time monitoring and data recording functions6The application of the online monitoring device in the ultra-high voltage power transformation equipment is more and more common.
SF6The on-line monitoring device is generally composed of signal acquisition, data processing and communication and display. The signal acquisition module mainly comprises a pressure sensor, a temperature sensor or a density sensor, and various sensors are usually arranged at a three-way valve opening of the extension of a breaker or a combined electrical apparatus tank body.
The data processing unit passes the SF according to the collected physical quantity6Gas equation to obtain SF at 20 deg.C6The pressure of the gas. The general gas equations are shown in equations (1) to (3):
P＝0.57×10-4ρT(1+B)-ρ2A (1)
A＝0.75×10-4(1-0.727×10-3ρ) (2)
B＝2.51×10-3ρ(1-0.84×10-3ρ) (3)
wherein P is SF6Pressure, p is SF6Density, T is ambient temperature. The relationship between the pressure P and the density ρ can be obtained as in the formula (4) by combining the formulas (1) to (3).
P＝(54.525-0.121T)ρ3+(0.143T-75)ρ2+0.057Tρ (4)
If the physical quantity measured by the detector is rho, the measured rho and T can be taken into formula (4) of 293.2K (20 ℃), and the GIS air chamber pressure value P at 20 ℃ can be obtained20；
If the physical quantities measured by the detector are P and T, iterative calculation is carried out according to the formulas (1) to (3) to obtain the gas density corresponding to the temperature, and then the pressure value of the GIS equipment converted to 20 ℃ can be obtained by carrying rho and T into 293.2K (20 ℃) and carrying the formula (4).
In theory, temperature compensation can counteract the ambient temperature change to SF6The influence of pressure change, namely, in a normal state, a pressure curve displayed by the online monitoring background is a straight line. In fact, because the sensor installation position of the online monitoring device is located at the three-way valve port extending out of the tank body, although the sensor is located inside the tank body, physical quantities such as pressure or density collected by the sensor are data at the three-way valve interface inside the tank body, the pressure or density at the local position is greatly influenced by the temperature change of the located position, even if the data obtained after gas equation conversion is not an approximately horizontal curve, the data is in a certain fluctuation rule, and particularly, the data has large fluctuation in the time period of large temperature difference or severe weather change in spring and autumn.
Typical SF of GIS equipment of certain extra-high voltage transformer substation6The online monitoring data is shown in fig. 1. The online monitoring device has continuous pressure data sampling intervals, high data precision and certain fluctuation regularity, and is suitable for prediction analysis.
2、SF6Analysis of influence factors of pressure changes
Based on the above analysis, due to SF6The on-line monitoring device sensor is arranged at the interface of a three-way valve of a GIS equipment shell, and the local temperature change of the position of a sampling point of the on-line monitoring device sensor influences the measurement result of the sensor. The invention starts from internal and external factors influencing the temperature change of the position of the sampling point and researches the SF of each factor6The influence of the pressure change of the device is monitored on line, and the analysis process is shown in figure 2.
2.1 external factors
External environment to SF6The influence of the temperature of the position where the sampling point of the on-line monitoring device is located is direct, such as sunshine insolation, cloudy and foggy rain or severe temperature change conditions and the like. Therefore, the invention firstly analyzes the factors such as the environmental temperature, the relative humidity and the weather type and the like to carry out the on-line monitoring on the device SF6Influence of pressure measurement results.
(1) Ambient temperature
The temperature of the sampling point is directly influenced by the temperature of the environment, and further the measuring result of the sensor of the on-line monitoring device is influenced, so that the environment temperature is selected as an influencing factor to be analyzed and is represented by a symbol T.
(2) Relative humidity
Relative humidity refers to the percentage of the water vapor pressure in air to the saturated water vapor pressure, and is indicated by the symbol RH. The local temperature change conditions of the object under different humidities are different, namely the humidity influences the temperature change of the position where the sensor is located, so that the relative humidity is selected as the influence SF6The factor of the pressure.
(3) Wind speed
The wind speed is the flow speed of the air and is denoted by the symbol F. Considering the effect of wind speed on the local temperature variation of the tank, wind speed is also referred to herein as the effect SF6One of the factors for monitoring the pressure variation of the device on line is studied,
(4) weather type
Because the positions of the sensors under different weather types (cloudy, sunny, rainy and snowy) receive different radiation degrees, the temperature sampling effect of the sensors is influenced, and the weather type is used as the influence SF6One of the factors for monitoring the pressure change of the device on line was studied. In order to carry out quantitative analysis, a fuzzy set theory method is adopted for weather type processing, a fuzzy function is used for representing weather characteristic values, the membership degree of each weather characteristic value is represented by a symbol X, and each weather characteristic value is shown in a table 1.
TABLE 1 weather eigenvalue membership
(5) Rate of change of temperature
The temperature change rate of the invention refers to the temperature change in a unit time interval, and because the influence of the intensity of the temperature change on the temperature sampling value of the sensor is also not negligible, the invention analyzes the environmental temperature change condition independently as an influence factor, and is expressed by a symbol delta theta, and the calculation formula is as follows:
wherein, T1For the temperature value at the current sampling moment, T0Is the temperature value at the last sampling moment, and t is the sampling time interval and has the unit of min.
2.2 internal factors
The temperature rise of the inner conductor of the electrical equipment is generated by current, so that the temperature rise in the GIS gas chamber is caused (the operation temperature rise can reach 65K), and further, the SF is treated6The temperature at the location of the on-line monitoring device sensor has an effect. Therefore, the invention selects the magnitude of the conductor current in the GIS equipment as SF6The influencing factor of the pressure change is indicated by symbol I.
2.3 correlation analysis
To calculate each influence factor pair SF6And the influence degree of the pressure change is obtained by selecting a typical seasonal online monitoring pressure value of operating equipment in an extra-high voltage station, corresponding external natural environment data and internal conductor current data and carrying out correlation calculation. The data sampling interval is 15 minutes, the total is 1800 groups, and the partial sample data condition is shown in the table 2.
TABLE 2 partial sample data
Obtaining various influencing factors and SF6The correlation results of the online monitored pressure values are shown in fig. 3.
In practical application, if the obtained fuzzy correlation coefficient is in a small range, such as [ -0.1, 0.1 ]]Within the interval, the correlation is considered to be weak and can be disregarded. The absolute values of the correlation coefficients of the 6 influencing factors and the pressure selected by the invention are all larger than 0.1, and the 6 influencing factors and the SF provided by the invention can be considered6There is a correlation in the pressure of the on-line monitoring device.
Wherein the ambient temperature T and the conductor currents I and SF6The absolute values of the correlation coefficients of the online monitoring pressure P exceed 0.5, and the signs are negative, so that the environment temperature T and the conductor current I are in a significant negative correlation with the pressure P. This is because according to equation (4), SF changes when the temperature T at the on-line monitoring device sensor changes6With negative correlation between the on-line monitored pressure P and the temperature T (measured SF)6The gas density is 6.088kg/m3And (4) nearby fluctuation, the density value can be substituted into a correlation formula to calculate the relation between P and T). The temperature change rates Δ θ and P are real correlations, and the remaining influencing factors are micro correlations with P.
3. BP neural network algorithm
The BP neural network is the most widely used one of the artificial neural networks, is also called an error back propagation network, and has the advantages of strong learning capacity, good nonlinear mapping capacity, good fault tolerance and the like. During the algorithm run, the information travels forward and the error propagates backwards to modify the network. The algorithm core of the network is a first-order gradient method (steepest descent method), and the sum of squares of errors between the actual output value and the ideal output value of the neural network is minimized by optimizing the connection weight between layers.
A typical BP neural network consists of an input layer, a hidden layer and an output layer, and a network model thereof is shown in fig. 4. X is the input layer of the network, the number of network nodes is N, wi,jConnecting weights for the input layer and the hidden layer, b,jFor the hidden layer threshold, the number of hidden layer network nodes is M, wj,kConnecting weights for hidden layer and output layer, b,kIs the output layer threshold, Y is the network output layer, and the number of nodes is Q.
The output mathematical model of the neuron in the network is as follows:
wherein x is the neuron input, u is the neuron output, w is the weight value, and b is the neuron threshold.
If Sigmoid is selected for the excitation function, the mathematical model is as follows:
thus, the output of the jth neuron of the hidden layer is obtained as:
similarly, the k-th neuron output of the output layer can be obtained as follows:
the error between the actual output value and the expected value after the network training is as follows:
wherein, OkThe expected value for the kth output sample.
And (3) substituting the output relation of each layer into an error formula (10) to obtain the relation between the error and the weight of each layer, and gradually adjusting the weight of each layer according to the descending direction of the error gradient until the error E meets the requirement.
4. SF based on BP neural network6Pressure prediction and analysis
4.1 data selection and processing
Selected SF according to the invention6The pressure data is typical seasonal pressure data (the on-line monitoring device manufacturer is Simatoex, Switzerland, the model of a sensor product is trafag8774) collected by an A-phase air chamber on-line monitoring device of a 1000kV certain interval breaker normally operated by a certain extra-high voltage substation, and the internal current data of the breaker at the corresponding moment comes from monitoring records in the substation. The environmental data of the corresponding time is acquired through the China Meteorological bureau official network.
The sampling interval of all data is 15 minutes, the total amount of the data is 1800 groups in total, in order to improve the convergence rate, a mapminmax function is used for carrying out normalization processing on various data, and the value range of each variable is [ -1, 1 ].
4.2BP neural network model design
The invention adopts a 3-layer topological structure BP neural network, 6 input layer nodes are adopted, namely N is 6, and influence SF is respectively adopted6Monitoring the ambient temperature, the relative humidity, the weather type, the wind speed, the temperature change rate and the conductor current of the pressure of the device on line; and M is 1 output layer node. The transfer function of the hidden layer is set to tansig, the transfer function of the output layer is set to logsig, the training function is set to train lm, the learning rate is 0.1, and the target precision is 0.00001. The number of hidden layer nodes is selected according to empirical formula (11).
And a is an adjusting constant, the value is 1-10, and the hidden layer nodes of the network are finally determined to be 13 through multiple training comparisons, namely M is 13.
The BP neural network prediction model is shown in fig. 5.
4.3BP neural network training and testing
The method uses a BP neural network tool box in MATLAB for simulation, and selects 1260 groups of data with the total sample amount of 75 percent as training samples of the network; selecting 15% of the total amount of samples, namely 270 groups of data as test samples; the remaining 15%, i.e. 270 sets of data, are used for verification. By the time of iteration 26, the network prediction results reach the best state, as shown in fig. 6.
It can be seen from the regression analysis (fig. 7) that the simulated BP neural network has better training, verification and testing conditions, and the R value of the overall data is about 0.94. The obtained BP neural network overall prediction error (training data, verification and test data) is shown in figure 8, the maximum error value of 1800 groups of samples is not more than +/-0.008 MPa, the prediction accuracy is more than 98.5%, the prediction result precision is high, and the correctness of theoretical analysis is verified.
A statistical Q-Q plot of the prediction error of a BP neural network is shown in FIG. 9, where most of the data points fall approximately on a straight line in the first quadrant, and thus SF can be approximated6The pressure prediction error of the online monitoring device follows normal distribution.
The prediction error distribution histogram of the BP network is shown in fig. 10.
The mean value of the prediction error is estimated to be-5.53 multiplied by 10-6MPa, error mean 95% confidence interval [ -1.52 × 10-5，4.14×10-6]MPa。
The trained BP neural network model is used for simulation, the obtained result is shown in fig. 11, and the obtained prediction model can better track the actual change trend and effectively predict pressure fluctuation caused by the change of the external natural environment and the internal current.
4.4 predictive applications
The model SF is determined according to the operation instruction provided by the manufacturer and the operation and maintenance practice of the operation and maintenance personnel6SF caused by external environment change in daily operation process of on-line monitoring device6A pressure false alarm condition occurs from time to time. When the defect occurs, operation and maintenance personnel often cannot judge whether the pressure change is caused by air leakage of the air chamber or external environment change through online monitoring data. According to the analysis, the pressure error data predicted by the BP neural network model established by the invention approximately obeys normal distribution, and the error expectation value is-5.53 multiplied by 10 within 95% confidence interval-6MPa, so that an appropriate error threshold value P can be set0And measuring the effectiveness of the measurement error, and further accurately analyzing the actual state of the equipment. According to the analysis of the actual pressure data change range of the on-line monitoring device and the recording precision of the equipment, the prediction error threshold value P can be obtained0Is set to be 2 x 10-3MPa. When GIS equipment air chamber SF6The pressure P is lower than the alarm value PWWhen a low-pressure alarm signal is sent out, the equipment operation and maintenance personnel can analyze the air leakage defect through the following strategies:
firstly, the result P' predicted by the BP neural network is compared with the actual pressure value P.
1) If the difference between P' and P is substantially less than P within 2 consecutive hours (15 minutes is one sampling point, and 8 sampling points are used in total)0And then, the air chamber pressure data can be judged to be normal, the air chamber pressure change is caused by the external environment change, and the air leakage fault of the equipment does not occur.
2) If the difference between P' and P is basically larger than P within 2 continuous hours0And the P value shows a descending trend along with the time change, and the equipment can be judged to be air leakage.
SF mentioned above6The pressure reduction cause determination strategy is shown in table 3.
TABLE 3 GIS devices SF6Pressure drop discrimination strategy
For GIS equipment of different types and phases, a proper BP neural network can be trained according to the current data of the external environment and the internal conductor of the GIS equipment to predict pressure, and a proper pressure prediction error threshold value P is set0. When the equipment generates an ' air pressure low alarm ' signal, the air chamber SF can be subjected to continuous observation of the difference relation between the prediction result P ' and the actual pressure P6The pressure drop is identified as the cause.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (10)
1. SF based on BP neural network6The method for predicting the pressure of the online monitoring device is characterized by comprising the following steps of:
(1) data selection and processing: selecting pressure data acquired by an online monitoring device of a phase air chamber to be detected of an interval circuit breaker, internal current data of the interval circuit breaker at a corresponding moment and environment data at the corresponding moment;
(2) designing a BP neural network model;
(3) BP neural network training;
(4) BP neural network pair SF6And predicting and analyzing the pressure of the online monitoring device.
2. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that in the step (1), various data are normalized by using a mapminmax function, and the value range of each variable is [ -1, 1]。
3. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that the environmental data comprises environmental temperature, relative humidity, weather type, wind speed and temperature change rate.
4. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that in the step (1), the sampling interval is 15 minutes.
5. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that in the step (2), the BP neural network model adopts a 3-layer topological structure BP neural network, 6 input layer nodes are respectively environment temperature, relative humidity, weather type, wind speed, temperature change rate and conductor current; the number of output layer nodes is 1.
6. The BP neural network-based SF according to claim 16The method for predicting the pressure of the on-line monitoring device is characterized in that in the step (2), the number of hidden layer nodes is 13。
7. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that in the step (3), a BP neural network tool kit in MATLAB is used for simulation training.
8. The BP neural network-based SF according to claim 16The method for predicting the pressure of the online monitoring device is characterized in that in the step (4), an error threshold value P is set by using pressure error data obtained by predicting a BP neural network model0And measuring the effectiveness of the measurement error, and further analyzing the actual state of the equipment.
9. The BP neural network-based SF according to claim 86The method for predicting the pressure of the on-line monitoring device is characterized in that in the step (4), when the equipment generates an alarm signal of low air pressure, the air chamber SF can be subjected to continuous observation of the difference relation between the prediction result P' and the actual pressure P6The pressure drop is identified as the cause.
10. The BP neural network-based SF according to claim 96The method for predicting the pressure of the online monitoring device is characterized in that in the step (4), a result P' obtained by predicting the BP neural network is compared with an actual pressure value P: if the difference between P' and P is substantially less than P within 2 consecutive hours0If the air chamber pressure data is normal, the air chamber pressure change is caused by external environment change, and air leakage fault of equipment does not occur; if the difference between P' and P is substantially larger than P within 2 hours of the reaction0And the P value shows a descending trend along with the time change, and the equipment can be judged to be air leakage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010944016.3A CN112182956A (en) | 2020-09-09 | 2020-09-09 | SF based on BP neural network6Method for predicting pressure of online monitoring device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010944016.3A CN112182956A (en) | 2020-09-09 | 2020-09-09 | SF based on BP neural network6Method for predicting pressure of online monitoring device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112182956A true CN112182956A (en) | 2021-01-05 |
Family
ID=73920436
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010944016.3A Pending CN112182956A (en) | 2020-09-09 | 2020-09-09 | SF based on BP neural network6Method for predicting pressure of online monitoring device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112182956A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114044496A (en) * | 2021-11-15 | 2022-02-15 | 国网河北省电力有限公司电力科学研究院 | Sulfur hexafluoride quality-based purification method, device and terminal based on neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214579A (en) * | 2018-09-20 | 2019-01-15 | 山东省林业科学研究院 | Salt-soda soil stability prediction method and system based on BP neural network |
| US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
| CN110728401A (en) * | 2019-10-10 | 2020-01-24 | 郑州轻工业学院 | Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm |
-
2020
- 2020-09-09 CN CN202010944016.3A patent/CN112182956A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
| CN109214579A (en) * | 2018-09-20 | 2019-01-15 | 山东省林业科学研究院 | Salt-soda soil stability prediction method and system based on BP neural network |
| CN110728401A (en) * | 2019-10-10 | 2020-01-24 | 郑州轻工业学院 | Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 周正思，刘林，程鹏: ""基于遗传算法优化BP神经网络的GIS设备放电故障诊断"", 《电气开关》 * |
| 李海波，祁文治，张建军: ""SF_6高压电器设备气体密度的监测及误差分析"", 《电力安全技术》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114044496A (en) * | 2021-11-15 | 2022-02-15 | 国网河北省电力有限公司电力科学研究院 | Sulfur hexafluoride quality-based purification method, device and terminal based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113899968B (en) | Voltage transformer monitoring method | |
| CN105631578A (en) | Risk assessment-orientated modeling method of power transmission and transformation equipment failure probability model | |
| CN105512962A (en) | Method for comprehensively evaluating insulation status of gas insulated switchgear (GIS) | |
| CN110136023A (en) | Powerline ice-covering risk profile based on adaptive enhancing study | |
| CN107886171A (en) | A kind of circuit-breaker status inline diagnosis method and system based on PMU data | |
| CN110542879B (en) | Method and system for predicting operation performance variation trend of capacitor voltage transformer | |
| CN109164248A (en) | A kind of predicting model for dissolved gas in transformer oil method | |
| CN112182956A (en) | SF based on BP neural network6Method for predicting pressure of online monitoring device | |
| Zhang et al. | Research on estimating method for the smart electric energy meter’s error based on parameter degradation model | |
| Fu et al. | Modelling and prediction techniques for dynamic overhead line rating | |
| CN110619105A (en) | Power transmission line temperature estimation method based on quantity measurement and heat balance equation | |
| CN111091223A (en) | Distribution transformer short-term load prediction method based on Internet of things intelligent sensing technology | |
| CN117200223A (en) | Day-ahead power load prediction method and device | |
| CN112232597A (en) | Safety prediction method based on multivariate long-short term memory network remote detection | |
| CN115733258A (en) | Control method of all-indoor intelligent substation system based on Internet of things technology | |
| CN112989695B (en) | Switch cabinet state evaluation method considering importance of power grid nodes | |
| CN112858812B (en) | Lightning arrester service performance evaluation method under extreme complex environment | |
| Yang et al. | Prediction of top oil Temperature for oil-immersed transformers Based on PSO-LSTM | |
| CN113537338A (en) | Robust line parameter identification method based on LSTM neural network and improved SCADA data | |
| CN112525438A (en) | SF (sulfur hexafluoride)6Air leakage monitoring method and system for density relay | |
| Wu et al. | Residual life prediction of mining cables based on RBF neural network | |
| CN107742886B (en) | Prediction method for load peak simultaneous coefficient of thermoelectric combined system | |
| Zeng et al. | A condition evaluation method of isolation circuit breaker based on triangular fuzzy analytic hierarchy process | |
| Man et al. | Research on Overload Current Measurement of Intelligent Process of Miniature Circuit Breaker Based on BP Neural Network | |
| Wang et al. | CVT error evaluation method based on moving window-weighted principal component analysis |
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 |