Insulator flashover voltage prediction method and system
By fusing mean square error and cross entropy error functions in the BP neural network model and performing data preprocessing and reprocessing, the problem of poor adaptability of model structure in insulator flashover voltage prediction is solved, and prediction accuracy and adaptability are improved. Deviation is reduced and physical mechanism impact is taken into account.
Patent Information
- Application Number
- CN202411425828.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-14
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prediction of insulator flashover voltage, the network model structure is fixed, making it difficult to adapt to the data characteristics of different insulator types and environmental conditions, resulting in large deviations in the prediction value and neglecting the uncertainty of the prediction results and the impact of physical mechanisms.
The improved BP neural network model is adopted to fuse the mean square error function with the cross entropy error function, and combine it with the hidden layer function to adjust the network structure, perform data preprocessing and reprocessing to improve prediction accuracy.
The accuracy and adaptability of insulator flashover voltage prediction are improved, the deviation between the prediction results and the actual value is reduced, and the uncertainty of the prediction results and the impact of the physical mechanism are taken into account.
Smart Images

Figure CN118917501A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of flashover voltage prediction, and in particular to a method and system for predicting flashover voltage of an insulator. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] Insulators play an important role in supporting and insulating the power system, ensuring the safe transmission of electric energy. However, insulators may experience flashover, which is a gas discharge phenomenon along the surface of the insulator. When an insulator flashes over, it will cause power transmission interruption, seriously affecting the stability and reliability of the power system. In addition, flashover may also cause irreversible damage to the insulator itself, shortening its service life and increasing the maintenance cost of the power system.
[0004] As the power system develops towards high voltage, large capacity and long distance, the performance requirements for insulators are getting higher and higher. The accurate prediction of insulator flashover voltage becomes particularly important. The accurate flashover voltage prediction method can realize real-time monitoring and early warning of the insulator state, which helps the power system operation and maintenance personnel to take measures in advance. With the continuous development of neural network technology, the technology of flashover voltage prediction based on neural network is developing faster and faster. In the prior art, for example, patent application number CN201510537544.6 discloses a method for predicting insulator flashover voltage, which specifically discloses sampling of leakage current and humidity as the input of the model, and the flashover voltage as the output of the model, and establishes a generalized regression neural network model for flashover voltage prediction. The study found that the prediction of insulator flashover voltage is a complex task involving multiple influencing factors. The network model structure in the prior art is fixed. When dealing with tasks such as insulator flashover voltage prediction with variable data characteristics, the adaptability is poor and it is difficult to accurately fit. It is impossible to make effective structural adjustments according to the data characteristics of different insulator types, different operating data and different environmental conditions, resulting in a large deviation between the predicted value of flashover voltage and the actual value. In addition, the existing technology often directly uses the output value of the network model as the final prediction value, thereby ignoring the uncertainty of the prediction result and the influence of the physical mechanism of the insulator, resulting in a large error in the flashover voltage prediction value. Summary of the invention
[0005] In order to solve the above problems, the present invention proposes an insulator flashover voltage prediction method and system. The present invention selects a BP neural network model, improves the traditional BP neural network model, combines the error function with the hidden layer function, and corrects the output value of the model. The network structure can be adjusted according to different data characteristics, and the accuracy of the prediction is improved.
[0006] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for predicting an insulator flashover voltage, comprising: Acquire insulator data, wherein the insulator data includes insulator property data, historical operation data, and historical environment data, and pre-process the acquired insulator data; Constructing an improved BP neural network model, wherein the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and combining the error function with a hidden layer function to update weights and biases in the improved BP neural network model; training the improved BP neural network model, and using the trained improved BP neural network model as a flashover voltage prediction model; The self-attribute data, current operation data and current environment data of the insulator to be tested are obtained, input into the flashover voltage prediction model, the initial flashover voltage prediction result is output, the initial flashover voltage prediction result is reprocessed, and the final flashover voltage prediction result is obtained.
[0007] According to a further technical solution, the self-attribute data includes: insulator type data, size data and surface contamination degree data; the historical operation data includes: voltage data, current data and operation time data; the historical environment data includes: temperature data, humidity data and air pressure data.
[0008] A further technical solution is that the preprocessing is to preprocess the acquired insulator property data, historical operation data and historical environmental data using an improved normalization method; the improved normalization method is specifically: = ,in, In order to use the data preprocessed by the improved normalization method, is the standardized data of insulator data, The insulator data is obtained.
[0009] In a further technical solution, the error function of the improved BP neural network model is specifically: E= + (1- ) , where E is the error function of the improved BP neural network model, is the mean square error function, is the cross entropy error function, is the fusion coefficient.
[0010] A further technical solution is to combine the error function with the hidden layer function, and the specific formula after the combination is: = (1- );in, is the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer; N is the number of neurons in the output layer; is the output value of the lth neuron in the output layer; is the target output value corresponding to the lth neuron in the output layer; is the connection weight from the jth neuron in the hidden layer to the lth neuron in the output layer; is the input value of the i-th neuron in the input layer; is the output value of the jth neuron in the hidden layer.
[0011] Further technical solution, the reprocessing is specifically: the initial flashover voltage prediction result output by the flashover voltage prediction model is , obtain the critical electric field strength of the insulator to be tested and electric field inhomogeneity coefficient , recalculate the flashover voltage using the reprocessing formula, and perform weighted average of the recalculated flashover voltage and the flashover voltage prediction result output by the initial flashover voltage prediction model to obtain the final flashover voltage prediction result; the reprocessing formula is: = d; where is the recalculated flashover voltage, and d is the electrode spacing.
[0012] A further technical solution, the reprocessing is specifically as follows: calculating the uncertainty interval of the initial flashover voltage prediction result, and constructing a weight function according to the interval width of the uncertainty interval; obtaining a flashover voltage reference value using an electric field distribution formula; and weighted averaging the flashover voltage reference value and the initial flashover voltage prediction result output by the flashover voltage prediction model to obtain a final flashover voltage prediction result.
[0013] In a second aspect, the present invention provides an insulator flashover voltage prediction system, comprising: The data acquisition and preprocessing module is configured to: acquire the insulator's own attribute data, historical operation data, and historical environmental data, and preprocess the acquired insulator's own attribute data, historical operation data, and historical environmental data; The prediction model building module is configured to: build an improved BP neural network model, the improved BP neural network model includes a fusion layer, an input layer, a hidden layer and an output layer in sequence; the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and the error function is combined with the hidden layer function to update the weights and biases in the improved BP neural network model; train the improved BP neural network model, and use the trained improved BP neural network model as a flashover voltage prediction model; The flashover voltage prediction module is configured to: obtain the self-property data, current operation data and current environmental data of the insulator to be tested, input them into the flashover voltage prediction model, output the initial flashover voltage prediction result, reprocess the initial flashover voltage prediction result, and obtain the final flashover voltage prediction result.
[0014] Compared with the prior art, the present invention has the following advantages and beneficial effects: The present invention selects the BP neural network model and improves the traditional BP neural network model, fuses the mean square error function and the cross entropy error function as the error function of the improved BP neural network model, and combines the error function with the hidden layer function. The problem of network model parameter optimization is solved. By continuously adjusting the weights and biases, the present invention can adjust the structure of the network model according to the data characteristics of different insulator types, different operating data and different environmental conditions. When processing the task of predicting the flashover voltage of the insulator with changeable data characteristics, the adaptability is strong and accurate fitting can be achieved, which reduces the deviation between the final prediction result and the actual value.
[0015] The present invention takes into account the uncertainty of the prediction results and the influence of the physical mechanism of the insulator. The output value of the network model is not directly used as the final prediction result, but the output value of the network model is reprocessed, thereby further improving the accuracy of the prediction result. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.
[0017] Figure 1 It is a flow chart of the method of embodiment 1 of the present invention. DETAILED DESCRIPTION
[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0019] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0020] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0021] Embodiment 1 like Figure 1 As shown, this embodiment proposes a method for predicting insulator flashover voltage, which specifically includes the following steps: S1: Acquire insulator data, wherein the insulator data includes insulator property data, historical operation data, and historical environment data, and pre-process the acquired insulator data; S2: constructing an improved BP neural network model, wherein the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and combining the error function with a hidden layer function to update weights and biases in the improved BP neural network model; training the improved BP neural network model, and using the trained improved BP neural network model as a flashover voltage prediction model; S3: Obtain the self-property data, current operation data and current environmental data of the insulator to be tested, input them into the flashover voltage prediction model, output the initial flashover voltage prediction result, reprocess the initial flashover voltage prediction result, and obtain the final flashover voltage prediction result.
[0022] Among them, in step S1, the acquired insulator attribute data includes but is not limited to insulator type data (disc insulator, column insulator, etc.), size data (height, length, diameter, etc.) and surface contamination degree data; historical operation data includes but is not limited to voltage data, current data and operation time data; historical environmental data includes but is not limited to temperature data, humidity data and air pressure data.
[0023] After the above insulator data is obtained, it needs to be preprocessed. The traditional preprocessing method is to process the data using the normalization method, but this method ignores the physical meaning of the data itself. In this embodiment, the preprocessing method is to preprocess the acquired insulator attribute data, historical operation data and historical environment data using the improved normalization method. The improved normalization method is specifically: = ,in, In order to use the data preprocessed by the improved normalization method, is the standardized data of insulator data, To obtain the insulator data, for example, to pre-process the obtained voltage data, assume is the standardized voltage or design voltage of the insulator, To obtain the current voltage data of the insulator, use = For another example, preprocess the acquired dimension data. Taking the column insulator as an example, assume is the standardized height or design height of the insulator, To obtain the current height data of the insulator, use = For example, the acquired insulator type data is preprocessed. Different types of insulators in this field have different structural coefficients, such as disc insulators, column insulators, etc., and the structural coefficients are different. Assume is the standardized structural factor or design structural factor of the insulator, To obtain the insulator type data, use = Preprocessing; Similarly, the temperature and humidity in the historical environmental data have insulated standardized data, which can be preprocessed using the improved normalization method mentioned above. This method makes the data no longer cold data, but data with physical meaning, which is more conducive to the subsequent prediction process and easier to explain the results.
[0024] In step S2, the improved BP neural network model includes an input layer, a hidden layer and an output layer. In this embodiment, the error function of the improved BP neural network model combines the mean square error function and the cross entropy error function. Specifically: the mean square error function = , =- , N is the number of neurons in the output layer, and the error function of the improved BP neural network model is E= + (1- ) , where E is the error function of the improved BP neural network model, is the mean square error function, is the cross entropy error function, is the fusion coefficient, ∈ , Used to control the mean square error function and the cross entropy error function The relative importance of in the error function E is adjusted by The value of can balance the effects of the two error functions in network training.
[0025] In this embodiment, the improvement of the BP neural network model is to combine the above error function with the hidden layer function to update the weights and biases in the improved BP neural network model. The specific combination process is: to obtain the mean square error function The partial derivative of the hidden layer output is as follows: First, = ,in, is the output value of the lth neuron in the output layer, followed by = , is the output value of the jth neuron in the hidden layer, where = + ,in, is the output layer bias, H is the number of hidden layer neurons, so = ,but = , and then find The partial derivative of the hidden layer output, = . Since E = + (1- ) , = ,final, = ,in =g( + ), where I is the number of neurons in the input layer, g() is a common activation function, and in this embodiment, g(z)= , = (1- ), so the specific formula after combination is: = (1- );in, is the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer; N is the number of neurons in the output layer; is the output value of the lth neuron in the output layer; is the target output value corresponding to the lth neuron in the output layer; is the connection weight from the jth neuron in the hidden layer to the lth neuron in the output layer; is the input value of the i-th neuron in the input layer; is the output value of the jth neuron in the hidden layer. The mean square error function and the cross entropy error function are fused as the error function of the improved BP neural network model, and the error function is combined with the hidden layer function. The problem of network model parameter optimization is solved. By continuously adjusting the weights and biases, the present invention can adjust the structure of the network model according to the data characteristics of different insulator types, different operating data and different environmental conditions. When processing the task of predicting the flashover voltage of the insulator with changeable data characteristics, the adaptability is strong and accurate fitting can be achieved, which reduces the deviation between the final prediction result and the actual value.
[0026] In step S3, the initial flashover voltage prediction result is reprocessed to obtain the final flashover voltage prediction result, specifically: In one embodiment, considering the influence of the physical mechanism of the insulator, the reprocessing is: Assuming that the initial flashover voltage prediction result output by the flashover voltage prediction model is , obtain the critical electric field strength of the insulator to be tested and electric field inhomogeneity coefficient The critical electric field strength under the current insulator electrode structure can be calculated based on the electric field distribution model. and electric field inhomogeneity coefficient , the existing technology can be realized by recalculating the flashover voltage using the reprocessing formula, and performing weighted average of the recalculated flashover voltage and the flashover voltage prediction result output by the initial flashover voltage prediction model to obtain the final flashover voltage prediction result; wherein the reprocessing formula is: = d; where is the recalculated flashover voltage, d is the electrode spacing. Taking high-voltage transmission lines as an example, the insulator is suspended between the tower and the conductor. The length of the insulator plus the distance from the connection point between the tower and the conductor to the insulator installation point together constitute the distance d between the electrodes (conductor-tower).
[0027] In another embodiment, the influence of uncertainty is further processed as follows: Calculate the initial flashover voltage prediction result The uncertainty interval is constructed according to the interval width of the uncertainty interval. The flashover voltage reference value is obtained using the electric field distribution formula. It should be noted that the insulator between the plate electrodes can be approximately regarded as a parallel plate capacitor, and the flashover voltage reference value can be obtained by referring to the electric field formula of the parallel plate capacitor. , where the electric field strength is the breakdown electric field strength of the insulator. For insulators between non-flat electrodes, such as some cylindrical insulators, refer to the electric field strength formula of cylindrical electrodes, where the electric field strength is still the breakdown electric field strength of the insulator. The flashover voltage reference value and the initial flashover voltage prediction result output by the flashover voltage prediction model are weighted averaged, that is, =w + (1-w) , and obtain the final flashover voltage prediction result.
[0028] Embodiment 2 This embodiment provides an insulator flashover voltage prediction system, which specifically includes the following modules: The data acquisition and preprocessing module is configured to: acquire the insulator's own attribute data, historical operation data, and historical environmental data, and preprocess the acquired insulator's own attribute data, historical operation data, and historical environmental data; The prediction model building module is configured to: build an improved BP neural network model, the improved BP neural network model includes a fusion layer, an input layer, a hidden layer and an output layer in sequence; the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and the error function is combined with the hidden layer function to update the weights and biases in the improved BP neural network model; train the improved BP neural network model, and use the trained improved BP neural network model as a flashover voltage prediction model; The flashover voltage prediction module is configured to: obtain the self-property data, current operation data and current environmental data of the insulator to be tested, input them into the flashover voltage prediction model, output the initial flashover voltage prediction result, reprocess the initial flashover voltage prediction result, and obtain the final flashover voltage prediction result.
[0029] The implementation of the specific modules in this embodiment refers to the steps of an insulator flashover voltage prediction method described in Embodiment 1, and will not be described in detail here.
[0030] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.
Claims
1. A method for predicting insulator flashover voltage, characterized in that: include: Acquire insulator data, wherein the insulator data includes insulator property data, historical operation data, and historical environment data, and pre-process the acquired insulator data; Constructing an improved BP neural network model, wherein the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and combining the error function with a hidden layer function to update weights and biases in the improved BP neural network model; Training the improved BP neural network model, and using the trained improved BP neural network model as a flashover voltage prediction model; The self-attribute data, current operation data and current environment data of the insulator to be tested are obtained, input into the flashover voltage prediction model, the initial flashover voltage prediction result is output, the initial flashover voltage prediction result is reprocessed, and the final flashover voltage prediction result is obtained.
2. The insulator flashover voltage prediction method according to claim 1, characterized in that: The self-attribute data includes: insulator type data, size data and surface contamination degree data; the historical operation data includes: voltage data, current data and operation time data; the historical environment data includes: temperature data, humidity data and air pressure data.
3. The insulator flashover voltage prediction method according to claim 1, characterized in that: The preprocessing is to preprocess the acquired insulator property data, historical operation data and historical environment data using an improved normalization method; the improved normalization method is specifically: = ,in, In order to use the data preprocessed by the improved normalization method, is the standardized data of insulator data, The insulator data is obtained.
4. The insulator flashover voltage prediction method according to claim 1, characterized in that: The error function of the improved BP neural network model is specifically: E= + (1- ) , where E is the error function of the improved BP neural network model, is the mean square error function, is the cross entropy error function, is the fusion coefficient.
5. The insulator flashover voltage prediction method according to claim 1, characterized in that: The error function is combined with the hidden layer function, and the specific formula after the combination is: = (1- );in, is the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer; N is the number of neurons in the output layer; is the output value of the lth neuron in the output layer; is the target output value corresponding to the lth neuron in the output layer; is the connection weight from the jth neuron in the hidden layer to the lth neuron in the output layer; is the input value of the i-th neuron in the input layer; is the output value of the jth neuron in the hidden layer.
6. The insulator flashover voltage prediction method according to claim 1, characterized in that: The reprocessing is specifically as follows: the initial flashover voltage prediction result output by the flashover voltage prediction model is , obtain the critical electric field strength of the insulator to be tested and electric field inhomogeneity coefficient , recalculate the flashover voltage using the reprocessing formula, and perform weighted average of the recalculated flashover voltage and the flashover voltage prediction result output by the initial flashover voltage prediction model to obtain the final flashover voltage prediction result; the reprocessing formula is: = d; where is the recalculated flashover voltage, and d is the electrode spacing.
7. The insulator flashover voltage prediction method according to claim 1, characterized in that: The reprocessing specifically includes: calculating the uncertainty interval of the initial flashover voltage prediction result, and constructing a weight function according to the interval width of the uncertainty interval; The flashover voltage reference value is obtained using the electric field distribution formula; The flashover voltage reference value and the initial flashover voltage prediction result output by the flashover voltage prediction model are weighted averaged to obtain the final flashover voltage prediction result.
8. An insulator flashover voltage prediction system, characterized in that: include: The data acquisition and preprocessing module is configured to: acquire the insulator's own attribute data, historical operation data, and historical environmental data, and preprocess the acquired insulator's own attribute data, historical operation data, and historical environmental data; The prediction model building module is configured to: build an improved BP neural network model, wherein the error function of the improved BP neural network model includes a mean square error function and a cross entropy error function, and combine the error function with a hidden layer function to update the weights and biases in the improved BP neural network model; Training the improved BP neural network model, and using the trained improved BP neural network model as a flashover voltage prediction model; The flashover voltage prediction module is configured to: obtain the self-property data, current operation data and current environmental data of the insulator to be tested, input them into the flashover voltage prediction model, output the initial flashover voltage prediction result, reprocess the initial flashover voltage prediction result, and obtain the final flashover voltage prediction result.
9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the insulator flashover voltage prediction method as described in any one of claims 1 to 7 are implemented.
10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the insulator flashover voltage prediction method as described in any one of claims 1 to 7 are implemented.
Citation Information
Patent Citations
Equivalent salt deposit density (ESDD) prediction and early warning system for power insulator surface contaminant
CN106405352A
Regional insulator group pollution flashover risk assessment method and system based on meteorological factors
CN112116176A
Composite insulator abnormal heating defect diagnosis method based on deep learning
CN113469950A
Shallow sea layered seabed earth sound parameter inversion method based on BP neural network model
CN115841076A
Insulator leakage current prediction method and device, equipment and storage medium
CN118245823A
