CN114492199A - Method and system for analyzing performance of voltage transformer, electronic equipment and medium - Google Patents
Method and system for analyzing performance of voltage transformer, electronic equipment and medium Download PDFInfo
- Publication number
- CN114492199A CN114492199A CN202210138807.6A CN202210138807A CN114492199A CN 114492199 A CN114492199 A CN 114492199A CN 202210138807 A CN202210138807 A CN 202210138807A CN 114492199 A CN114492199 A CN 114492199A
- Authority
- CN
- China
- Prior art keywords
- weight vector
- voltage transformer
- data set
- neural network
- algorithm
- 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
- 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
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a method, a system, electronic equipment and a medium for analyzing the performance of a voltage transformer, relates to the technical field of electrical measurement, solves the problem that the prior art can not effectively realize high-precision trend prediction of the voltage transformer by capturing time series information and nonlinear characteristics of performance change, and has the technical scheme that: the method comprises the steps of carrying out regression calculation on a test data set based on a dendritic neural network regression model constructed by an SFSMS algorithm, calculating a weight vector at the output end of the model and the last training data set to obtain a predicted value, comparing the predicted value with the test data set, obtaining the difference between the predicted value and a real test data set based on the difference value, and inputting data of a voltage transformer into the dendritic neural network regression model to output a value which can be used for accurately evaluating the operation state of the capacitor voltage transformer in a certain time period if the difference is small and the obtained predicted value is close to a real value.
Description
Technical Field
The invention relates to the technical field of electrical measurement of voltage transformers, in particular to a method and a system for analyzing the performance of a voltage transformer, electronic equipment and a medium.
Background
A voltage transformer is an important electrical device in an electrical network for converting a high voltage into a low voltage for use by metering instruments and protective equipment. Among them, a Capacitor Voltage Transformer (CVT) is widely used in 110kV to 500kV power grids, gradually replacing an electromagnetic voltage transformer, due to its advantages of high impact insulation strength, simple manufacture, small size, light weight, and remarkable economical efficiency.
The existing online monitoring system and method aiming at the capacitor voltage transformer in the power grid system can obtain real-time insulation and metering parameters of equipment and reliably and stably monitor the running performance of the equipment. However, the current machine learning method for online monitoring is only limited to a simple regression model, and cannot effectively realize high-precision trend prediction by capturing time series information and nonlinear characteristics of performance change, and some situations of low performance change trend prediction precision exist, so that the method cannot be effectively applied to actual engineering application.
Disclosure of Invention
The invention aims to provide a method, a system, electronic equipment and a medium for analyzing the performance of a voltage transformer, which are used for solving the problem that the prior art can not effectively realize high-precision trend prediction by capturing time series information and nonlinear characteristics of performance change, so that some situations of low performance change trend prediction precision exist, and the performance change trend prediction precision can not be effectively applied to actual engineering. The invention provides a method for analyzing the performance of a voltage transformer, which aims to improve the precision and efficiency of predicting the operation performance variation trend of the voltage transformer.
The technical purpose of the invention is realized by the following technical scheme:
in a first aspect, the present invention provides a method for analyzing performance of a voltage transformer, including the following steps:
s1, collecting environmental parameters and operating parameters in an area where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise the first N-1 training data sets and the Nth testing data set;
s2, constructing a dendritic neural network regression model according to the SFSMS algorithm;
s3, initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
s4, updating the first weight vector according to the loss weight value to obtain a second individual weight vector;
s5, setting i to i +1, determining whether i is equal to N, if so, executing step S6, otherwise, repeating steps S3-S5, iteratively updating the second weight vector until i is equal to N, and outputting a third weight vector representing the individual optimal weight result;
and S6, calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
Compared with the prior art, the invention firstly provides a trend prediction method of dendritic neural regression based on scale-free material state search, which can enhance the regression capability of a model by considering the thickness of dendritic branches (the thickness is related to signal intensity). The dendritic neural network regression model consists of four layers of networks: synapse layer, dendron layer, membrane layer and cell layer. Due to the nonlinear characteristic of a synapse layer and the plasticity of a dendron layer, the dendritic neural network regression model has strong capability of fitting a complex nonlinear function, the accuracy of trend prediction is improved, and in addition, the weight of the dendritic neural network regression model also has a complex and large search space, so a new scaleless material state search (SFSMS) algorithm is adopted, and the algorithm combines a material state search (SMS) algorithm with a scaleless local search method to optimize the neural network structure of the dendritic neural network regression model. The SMS algorithm has strong searching capability, can effectively avoid local optimization, and improves the efficiency and generalization capability of the model. Inputting a sample data set for collecting environmental parameters and operating parameters of a region where the voltage transformer is located into a neural network regression model constructed by a new scale-free material state search (SFSMS) algorithm, calculating the parameters in the sample data set by the model to finally obtain a third weight vector representing optimal parameter performance, calculating a predicted value of an Nth test data set based on the third weight vector and an N-1 th training data set, comparing the predicted value with test data of an actual voltage transformer, and obtaining the accuracy of the predicted value obtained by the analysis method according to the difference value between the predicted value and the test data, so that the analysis method can be effectively applied to prediction of the operating performance change trend of the capacitor voltage transformer.
Further, the environmental parameters of the sample data set comprise average atmospheric temperature and average atmospheric humidity in a day in an area where the voltage transformer is located, and the operation parameters comprise performance parameter ratio error values calculated according to a secondary voltage effective value of the voltage transformer;
and respectively setting parameters of a synapse layer and a cell layer of the dendritic neural network regression model as alpha and v according to an SFSMS algorithm, wherein the number of branches of a dendron layer is M, the size of a population is N, and the maximum iteration number is Ita.
Further, obtaining a single first weight vector by applying the single training data set to a dendritic neural network regression model, wherein the feature quantity of the single first weight vector is the number of days multiplied by the sum of the parameters in the sample data set;
and calculating the length of a first weight vector of the dendritic neural network regression model according to the characteristic quantity and the branch number of the dendrobe layer.
Further, constructing a loss function of the SFSMS algorithm according to the first weight vector and the actual value of the performance of the voltage transformer.
Further, a population set of the dendritic neural network regression model is constructed according to the first weight vectors and an SFSMS algorithm, wherein the length of each first weight vector in the population set represents the weight and the threshold value of a training data set obtained through calculation of the dendritic neural network regression model.
Further, initializing an SFSMS algorithm iteration parameter t as 1, updating parameters of a synapse layer and a cell layer of the dendritic neural network regression model according to the SFSMS algorithm, and calculating the loss weight value of the loss function corresponding to the first weight vector in the SFSMS algorithm;
updating the first weight vector in the population set according to the loss weight value to obtain a second weight vector of an individual;
and updating the second weight vector until the iteration parameter t is greater than the maximum iteration times to obtain a third weight vector representing the individual optimal weight result.
Furthermore, the predicted value and the test data set are compared to compare the prediction precision of the performance of the voltage transformer, wherein the element of the corresponding position of the predicted value corresponding to the third column vector of the test data set after being flattened into one dimension is a performance parameter ratio error value of several days after prediction.
In a second aspect, the present invention provides a system for analyzing the performance of a voltage transformer, comprising:
the voltage transformer testing system comprises a data acquisition unit, a data acquisition unit and a data analysis unit, wherein the data acquisition unit is used for acquiring environmental parameters and operating parameters in an area where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise first N-1 training data sets and Nth testing data sets;
the model building unit is used for building a dendritic neural network regression model according to the SFSMS algorithm;
the first calculation unit is used for initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
the second calculating unit is used for updating the first weight vector according to the loss weight value to obtain a second weight vector of an individual;
a judging unit, configured to set i to i +1, judge whether i is equal to N, if so, execute a third calculating unit, otherwise, repeatedly execute the first calculating unit and the second calculating unit, iteratively update the second weight vector until i is equal to N, and output a third weight vector representing an individual optimal weight result;
and the third calculating unit is used for calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
In a third aspect, the present invention further provides an electronic device, including: the system comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus; the memory for storing a computer program; the processor is configured to execute the program stored in the memory to implement the method for analyzing the performance of the voltage transformer according to the first aspect.
In a fourth aspect, the present invention also provides a computer-readable storage medium for storing computer-readable instructions, which when executed perform the operations of the method for analyzing the performance of the voltage transformer provided in the first aspect.
Compared with the prior art, the invention has the following beneficial effects:
the invention firstly provides a trend prediction method of dendritic neural regression based on scale-free material state search, which can enhance the regression capability of a model by considering the thickness of dendritic branches (the thickness is related to signal intensity). The dendritic neural network regression model consists of four layers of networks: synapse layer, dendron layer, membrane layer and cell layer. Due to the nonlinear characteristic of a synapse layer and the plasticity of a dendron layer, the dendritic neural network regression model has strong capability of fitting a complex nonlinear function, the accuracy of trend prediction is improved, and in addition, the weight of the dendritic neural network regression model also has a complex and large search space, so a new scaleless material state search (SFSMS) algorithm is adopted, and the algorithm combines a material state search (SMS) algorithm with a scaleless local search method to optimize the neural network structure of the dendritic neural network regression model. The SMS algorithm has strong searching capability, can effectively avoid local optimization, and improves the efficiency and generalization capability of the model. Inputting a sample data set for collecting environmental parameters and operating parameters of a region where the voltage transformer is located into a neural network regression model constructed by a new scale-free material state search (SFSMS) algorithm, calculating the parameters in the sample data set by the model to finally obtain a third weight vector representing optimal parameter performance, calculating a predicted value of an Nth test data set based on the third weight vector and an N-1 th training data set, comparing the predicted value with test data of an actual voltage transformer, and obtaining the accuracy of the predicted value obtained by the analysis method according to the difference value between the predicted value and the test data, so that the analysis method can be effectively applied to prediction of the operating performance change trend of the capacitor voltage transformer.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a schematic flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a block diagram of a system framework according to an embodiment of the present invention;
fig. 3 is a block diagram of an electronic device according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and the accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not used as limiting the present invention.
It should be noted that the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
The first embodiment is as follows:
as shown in fig. 1, an embodiment of the present application provides a method for analyzing performance of a voltage transformer, including the following steps:
s1, collecting environmental parameters and operating parameters in the region where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise the first N-1 training data sets and the Nth testing data set.
Specifically, data of a training data set and a test data set acquired by the voltage transformer are from various parameter records of a capacitive voltage mutual sensor of a real line, and are grouped according to a time window sequence by taking a day as a unit, and each group is used as a sample data set.
S2, constructing a dendritic neural network regression model according to the SFSMS algorithm.
Specifically, in S2 of the first embodiment, the SFSMS algorithm is composed of an SMS algorithm and a scale-free local search method, and a dendritic neural network model (i.e., DNR) is constructed by using the SFSMS algorithm, where the DNR includes a stratum network, which is a synapse layer, a dendron layer, a membranous layer, and a cellular layer. The synapse layer is the entry point of the model for receiving the input signal. The signals received by the synapse layer are processed by the activation function and then flow to all branches of the dendron layer. Each branch of the dendron layer collects all the signals in the corresponding branch and sends them to the membrane layer. The membrane layer receives signals from all branches of the dendron layer and integrates them for transmission to the cell layer. And finally, processing the signals by the soma layer through a Sigmoid function and outputting the signals. The calculation of DNR mimics the process by which biological neurons transmit information. The network layers are described in detail as follows:
and (3) a synapse layer: the synaptic layer mimics the synaptic part of the nervous system. It is the entrance of a neuron that receives signals from external inputs. Given data X, comprising N sets of time-sequential samples, each sample comprising l signals, the signals received by the synapse layer are processed by the following equation:
wherein x isiIs the ith input signal, DijA value representing the transfer of the ith synapse to the jth dendrite. α is a normal number parameter of the synapse layer. w is aijAnd θ are two variable parameters for different tasks.
And (3) outburst layer: the dendron layer is responsible for gathering the signals emitted by synapses distributed over each branch. The non-linear relationship between these signals, which is thought to play an important role in the neural information processing of certain sensory systems (e.g., the visual and auditory systems) in biological networks, is described by multiplication in DNR, which can be expressed as:
wherein M isjRepresenting the output value of the jth dendrite branch.
Film layer: integration of the membrane layer results from the signal from all branches of the dendrites. The integral operation is realized by a discrete summation form, and the specific formula is
Wherein mujRepresents the strength of each dendritic branch, where J is the number of dendritic branches and S represents the input to the cell layer. Mu.sjIs a control parameter that distinguishes whether the model is specifically for a regression task or a classification task. In the classification task, it is a constant 1. In the regression task, the regression task is carried out,μjis a variable parameter that can be changed to adapt to different tasks to better deal with regression problems.
Cell layer: in the cell layer, Sigmoid function is used as the activation function. When the signal from the membrane layer exceeds a threshold, the cell body is excited. The process can be defined as
Where R is the output of the cell layer and α and v are two normal number parameters.
(1) Defining the total size N of a sample, the dimension N, the maximum iteration number Ita and the number M of initial nodes0。
(2) Initializing a population
Set of direction vectors
The maximum iteration count Ita is 1000, the state counting phase is 1, and the current iteration count m is 1.
(3) Confirming each parameter through the value of phase in different stages, which is as follows:
(4) calculating fitness function of population
Setting the current best individual as
According to the formula
Computing a new set of direction vectors
According to a formula
Obtaining a set of velocity vectors
(5) Then using the formula
To update the population. By the formula
To calculate a threshold k and then to calculate the distance between each individual, and if the distance is less than k, then using the formula
To exchange the individual direction vectors. Obtain a [0,1 ]]A random number in between
According to the formula
To perform random behavior.
(6) Using BA algorithm to generate scale-free network, numbering each node, and calculating fitness function of population again
The individuals are then ranked according to fitness, and each individual is placed into a network node having the same number and rank. By the formula
Each volume is updated and the iteration number m is increased by one.
(7) Repeating the above steps (4) - (5) PD times according to different stages.
(8) And (4) repeating the steps (3) - (7) for 3 times in total, and after each iteration, adding one to the phase in the counting stage.
(9) Finally outputting the best individual
And calculating the model through the nine steps of SFSMS, and obtaining the predicted value of the operation performance of the voltage transformer through the subsequent S3-S6.
S3, initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
and S4, updating the first weight vector according to the corresponding loss weight value to obtain a second weight vector of the individual.
S5, setting i to i +1, determining whether i is equal to N, if so, executing step S6, otherwise, repeating steps S3-S5, iteratively updating the second weight vector until i is equal to N, and outputting a third weight vector representing the individual optimal weight result;
and S6, calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
In steps S3-S6 in this embodiment, a test data set is subjected to regression calculation based on a dendritic neural network regression model constructed by an SFSMS algorithm, a weight vector is optimized according to iterative update capability of the model, a predicted value of N test data sets is obtained by calculating the weight vector at the output end of the model and the last test data set, the predicted value is compared with the N test data sets, a difference between the predicted value and a true test data set can be obtained based on the magnitude of the difference, and if the difference is small, it is determined that the obtained predicted value is close to the true value, a value output by inputting data of a voltage transformer into the dendritic neural network regression model constructed based on the SFSMS is used for evaluating the operating state of the capacitor voltage transformer in a certain time period.
The existing online monitoring system and method aiming at the capacitor voltage transformer in the power grid system can obtain real-time insulation and metering parameters of equipment and reliably and stably monitor the running performance of the equipment. However, the current machine learning method for online monitoring is only limited to a simple regression model, cannot effectively capture time series information and nonlinear characteristics, has low accuracy for predicting the performance change trend, and cannot achieve the purposes of accurately predicting and early warning the state of future equipment with high accuracy. The first embodiment of the present application proposes a trend prediction method of dendritic neural regression based on scale-free material state search, which can enhance the regression capability of the model by considering the thickness of the dendritic branches (the thickness is related to the signal intensity). The dendritic neural network regression model consists of four layers of networks: synapse layer, dendron layer, membrane layer and cell layer. Due to the nonlinear characteristic of a synapse layer and the plasticity of a dendron layer, the dendritic neural network regression model has strong capability of fitting a complex nonlinear function, the accuracy of trend prediction is improved, and in addition, the weight of the dendritic neural network regression model also has a complex and large search space, so a new scaleless material state search (SFSMS) algorithm is adopted, and the algorithm combines a material state search (SMS) algorithm with a scaleless local search method to optimize the neural network structure of the dendritic neural network regression model. Because the SMS algorithm has strong searching capability, local optimization can be effectively avoided, and therefore the efficiency and the generalization capability of the regression model are improved. Inputting a sample data set for collecting environmental parameters and operating parameters of a region where the voltage transformer is located into a neural network regression model constructed by a new scale-free material state search (SFSMS) algorithm, calculating the parameters in the sample data set by the model to finally obtain a third weight vector representing optimal parameter performance, calculating a predicted value of an Nth test data set based on the third weight vector and an N-1 th training data set, comparing the predicted value with test data of an actual voltage transformer, and obtaining the accuracy of the predicted value obtained by the analysis method of the invention, which is an actual value close to the operating performance of the voltage transformer, according to the magnitude of a difference value between the predicted value and the test data of the actual voltage transformer, so that the analysis method of the invention can be effectively applied to prediction of the operating performance change trend of the capacitor transformer.
In the conventional practical application, the power company mainly determines the operation state of the voltage transformer by means of power failure cycle verification, and power failure is required when the regulation is executed. In addition, the method needs the staff to carry equipment to the field and disassemble and assemble the primary lead of the tested voltage transformer, has the defects of large workload, low efficiency, untimely fault defect discovery and the like, and influences the safety operation of the electric power system and the fairness and justice of electric energy metering.
The accuracy of the predicted value obtained by the analysis method is close to the actual value of the operation performance of the voltage transformer, so that the operation state of the voltage transformer can be judged by a power company without a power failure period verification mode, and the detection on the performance of the voltage transformer can be realized only by collecting the operation parameters of the voltage transformer. In addition, according to the method, the staff does not need to carry equipment to rush to the site and disassemble and assemble the primary lead of the tested voltage transformer, so that the workload of the staff is greatly reduced, the efficiency is improved, and the safe operation of the electric power system and the normal metering of the electric energy metering are not influenced.
In yet another embodiment of the first embodiment of the present application, the environmental parameters of the sample data set include an average atmospheric temperature and an average atmospheric humidity in a day in an area where the voltage transformer is located, and the operating parameters include performance parameter ratio error values calculated according to a secondary voltage effective value of the voltage transformer;
parameters of a synaptic layer and a cellular layer of a dendritic neural network regression model are respectively given as alpha and v according to an SFSMS algorithm, the number of branches of a dendron layer is M, the size of a population is N, and the maximum iteration number is Ita.
Specifically, given L days of data, each day of data comprises average air temperature data T, average atmospheric humidity H and effective value U of secondary voltagecvtAnd calculating the error f of the obtained average performance parameter ratio. Wherein the content of the first and second substances,
and the average value of the effective values of the secondary voltages of the voltage transformers in one day is represented. The data sample of the first day may be denoted as S1＝[T1,H1,f1]In this way, the sample at the L-th day is SL＝[TL,HL,fL]Grouping the N groups of data according to the time window size of c days to finally obtain N groups of data, namely G ═ G1,G2,…,CNAnd where the relationship cxn ≦ L is satisfied. Dividing the data into two parts of data sets according to time sequence, and taking the first N-1 groups of data as a training data set GtrainAnd the last group is a test data set Gtest. The environment parameters comprise the average atmospheric temperature and the average atmospheric humidity in the area where the voltage transformer is located in one day, the air temperature difference of the area is mainly considered, so the average atmospheric temperature and the humidity of the atmosphere are solved, the operation parameters comprise performance parameter ratio error values obtained through calculation according to the effective value of the secondary voltage of the voltage transformer, and the parameters of the model are set for iterative calculation of a subsequent neural network regression model.
In yet another embodiment of the first embodiment of the present application, obtaining a single first weight vector by applying a single training data set to a dendritic neural network regression model, wherein the feature quantity of the single first weight vector is the number of days multiplied by the sum of the parameters in the sample data set;
and calculating the length of a first weight vector of the dendritic neural network regression model according to the characteristic quantity and the branch number of the dendrobe layer.
Specifically, parameters alpha and v of a synapse layer and a cell layer, the number M of branches of a tree conflict layer, the population size N and the maximum iteration number Ita are given. Defining a DNR calculation process, wherein the specific calculation process is shown in an introduction part of a dendritic neural network model, inputting a training sample to a DNR model to obtain an output O, defining the length D of a DNR weight vector, wherein the characteristic quantity I is 3 × c, namely three average parameter values of c days are contained in one sample, and D is 2MI + 1.
In yet another embodiment of the first embodiment of the present application, the loss function of the SFSMS algorithm is constructed based on the first weight vector and the actual value of the voltage transformer performance.
In particular, a loss function defining the SFSMS algorithm
In this embodiment, the predicted true value of the jth sample is the j +1 th sample, and E is an error loss representing the predicted and true output values, and the loss function of the SFSMS algorithm constructed in this embodiment performs error compensation on the calculated weight vector, so that the finally obtained predicted value is closer to the true value of the operation performance of the voltage transformer.
In yet another embodiment of the first embodiment of the present application, a population set of the dendritic neural network regression model is constructed according to the first weight vector and the SFSMS algorithm, wherein the length of each first weight vector in the population set represents the weight and the threshold value of a training data set calculated by the dendritic neural network regression model.
Specifically, the SFSMS algorithm is used for updating the parameters of the neural network regression model, and the population P in the algorithm is defined as { W }1,W2,…,WNIn which WiAnd representing an ith weight vector corresponding to the ith training data set, wherein each weight vector comprises D parameters and represents the weight and the threshold value obtained by DNR calculation of one training data set.
In another embodiment of the first embodiment of the present application, initializing an iteration parameter t ═ 1 of an SFSMS algorithm, updating parameters of a synapse layer and a cell layer of a dendritic neural network regression model according to the SFSMS algorithm, and calculating a loss weight value of a loss function corresponding to a first weight vector in the SFSMS algorithm;
updating the first weight vector in the population set according to the loss weight value to obtain a second weight vector of the individual;
and updating the second weight vector until the iteration parameter t is greater than the maximum iteration times to obtain a third weight vector representing the individual optimal weight result.
Specifically, a parameter t ═ 1 is defined and initialized, and the phase state parameters and related parameters are updated according to the SFSMS algorithm.
For each first weight vector WiInputting all training data sets into DNR to obtain corresponding output and calculating corresponding loss weight value of SFSMS
Updating the population P according to the loss weight values and obtaining an optimal weight vector (individual) WbestAnd updating t to t + 1.
Repeating the steps in the embodiment completes the iterative update of DNR until t is greater than the maximum number of iterations Ita, thereby outputting a third weight vector for the individual optimal weight result by the dendritic neural network regression model.
In another embodiment of the first embodiment of the present application, the prediction accuracy of the performance of the voltage transformer may be compared by comparing the predicted value with the test data set, where an element at a corresponding position where the predicted value corresponds to the third column vector of the test data set after being flattened into a one-dimensional form is a predicted performance parameter ratio error value for several days after prediction.
Example two:
based on the same concept, as shown in fig. 2, a second embodiment of the present application provides an analysis system for performance of a voltage transformer, and specific implementation of the system may refer to a description of a part of the embodiment of the method, and repeated parts are not described again, including:
the data acquisition unit 110 is configured to acquire environmental parameters and operating parameters in an area where the voltage transformer is located, and construct N sample data sets according to the environmental parameters and the operating parameters, where the N sample data sets include first N-1 training data sets and an nth test data set;
a model construction unit 120 for constructing a dendritic neural network regression model according to the SFSMS algorithm;
a first calculating unit 130, configured to initialize the number i of training data sets to 1, input the ith training data set into the dendritic neural network regression model for calculation, obtain a first weight vector corresponding to the ith training data set, and calculate a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
the second calculating unit 140 is configured to update the first weight vector according to the loss weight value, so as to obtain a second weight vector of an individual;
a determining unit 150, configured to set i to i +1, determine whether i is equal to N, if so, execute a third calculating unit, otherwise, repeatedly execute the first calculating unit 130 and the second calculating unit 140, iteratively update the second weight vector until i is equal to N, and output a third weight vector representing an individual optimal weight result;
and the third calculating unit 160 is configured to calculate a predicted value of the nth test data set according to the third weight vector and the (N-1) th training data set.
A second embodiment provides a system for analyzing performance of a voltage transformer, where in the second embodiment, the first calculating unit 130, the second calculating unit 140, the determining unit 150, and the third calculating unit 160 all perform regression calculation on a test data set based on a dendritic neural network regression model constructed by an SFSMS algorithm, optimize a weight vector according to an iterative updating capability of the model, calculate a weight vector at an output end of the model and a last test data set to obtain predicted values of N test data sets, compare the predicted values with the N test data sets, and obtain a difference between the predicted values and a real test data set based on a magnitude of the difference, where if the difference is small, the obtained predicted values are close to real values, a value output by inputting data of the voltage transformer into the dendritic neural network regression model constructed based on the SFSMS is used for evaluating an operating state of the capacitor voltage transformer for a certain period of time And (6) estimating.
The accuracy of the predicted value obtained by the analysis system based on the second embodiment of the application is an actual value close to the operation performance of the voltage transformer, so that the power company can judge the operation state of the voltage transformer without a power failure period verification mode, and the detection on the performance of the voltage transformer can be realized only by acquiring the operation parameters of the voltage transformer. In addition, according to the method, the staff does not need to carry equipment to rush to the site and disassemble and assemble the primary lead of the tested voltage transformer, so that the workload of the staff is greatly reduced, the efficiency is improved, and the safe operation of the electric power system and the normal metering of the electric energy metering are not influenced.
In another embodiment of the second embodiment of the present application, the data acquisition unit 110 includes a data acquisition subunit, where the environment parameters of the sample data set include an average atmospheric temperature and an average atmospheric humidity in a day in an area where the voltage transformer is located, and the operation parameters include performance parameter ratio error values calculated according to a secondary voltage effective value of the voltage transformer;
the model construction unit 120 includes a parameter setting unit, configured to set parameters of a synapse layer and a cell layer of the dendritic neural network regression model as α and v, respectively, according to the SFSMS algorithm, the number of branches of the dendron layer is M, the population size is N, and the maximum iteration number of the dendritic neural network regression model is Ita.
In another embodiment of the second embodiment of the present application, the first calculating unit 130 includes a first calculating subunit, configured to apply a single training data set to the dendritic neural network regression model to obtain a single first weight vector, where the feature quantity of the single first weight vector is a number of days multiplied by a sum of parameters in the sample data set;
and calculating the length of a first weight vector of the dendritic neural network regression model according to the characteristic quantity and the branch number of the dendrobe layer.
In another embodiment of the second embodiment of the present application, the first calculating subunit is further configured to construct a loss function of the SFSMS algorithm according to the first weight vector and the actual value of the performance of the voltage transformer.
In another embodiment of the second embodiment of the present application, the second calculating unit 140 includes a second calculating subunit, configured to construct a population set of the dendritic neural network regression model according to the first weight vectors and the SFSMS algorithm, where a length of each first weight vector in the population set represents a weight and a threshold value of one training data set calculated by the dendritic neural network regression model.
In another embodiment of the second embodiment of the present application, the determining unit 150 includes a determining subunit, configured to initialize an iteration parameter t of the SFSMS algorithm to 1, update parameters of a synapse layer and a cell layer of the dendritic neural network regression model according to the SFSMS algorithm, and calculate a loss weight value of a loss function corresponding to the first weight vector in the SFSMS algorithm;
updating the first weight vector in the population set according to the loss weight value to obtain a second weight vector of the individual;
and updating the second weight vector until the iteration parameter t is greater than the maximum iteration times to obtain a third weight vector representing the individual optimal weight result.
In another embodiment of the second embodiment of the present application, the analysis system further includes a comparison unit, configured to compare the predicted value with the test data set, so as to compare the prediction accuracy of the performance of the voltage transformer, where an element at a corresponding position where the predicted value corresponds to the third column vector of the test data set after being flattened into a one-dimensional is a performance parameter ratio error value for several days after prediction.
Example three:
based on the same concept, a third embodiment of the present application provides an electronic device, as shown in fig. 3, including: the processor 310, the communication interface 320, the memory 330 and the communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 are communicated with each other through the communication bus 340; the memory 330 for storing a computer program; the processor 310 is configured to execute the program stored in the memory 330, and implement the following steps: s1, collecting environmental parameters and operating parameters in an area where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise the first N-1 training data sets and the Nth testing data set; s2, constructing a dendritic neural network regression model according to the SFSMS algorithm; s3, initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector; s4, updating the first weight vector according to the loss weight value to obtain a second individual weight vector; s5, setting i to i +1, determining whether i is equal to N, if so, executing step S6, otherwise, repeating steps S3-S5, iteratively updating the second weight vector until i is equal to N, and outputting a third weight vector representing the individual optimal weight result; and S6, calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
Example four:
in a fourth aspect, the present invention further provides a computer readable storage medium for storing computer readable instructions, wherein the instructions, when executed, perform the operations of the method for analyzing the performance of a voltage transformer as provided in the first aspect. The content of the description of the analysis method of the performance of the voltage transformer is consistent with that of the first embodiment, and is not described in detail herein.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A method for analyzing the performance of a voltage transformer is characterized by comprising the following steps:
s1, collecting environmental parameters and operating parameters in an area where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise the first N-1 training data sets and the Nth testing data set;
s2, constructing a dendritic neural network regression model according to the SFSMS algorithm;
s3, initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
s4, updating the first weight vector according to the loss weight value to obtain a second individual weight vector;
s5, setting i to i +1, determining whether i is equal to N, if so, executing step S6, otherwise, repeating steps S3-S5, iteratively updating the second weight vector until i is equal to N, and outputting a third weight vector representing the individual optimal weight result;
and S6, calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
2. The method according to claim 1, wherein the environmental parameters of the sample data set include an average atmospheric temperature and an average atmospheric humidity in a day in an area where the voltage transformer is located, and the operating parameters include performance parameter ratio error values calculated according to an effective value of a secondary voltage of the voltage transformer;
and respectively setting parameters of a synapse layer and a cell layer of the dendritic neural network regression model as alpha and v according to an SFSMS algorithm, wherein the number of branches of a dendron layer is M, the size of a population is N, and the maximum iteration number is Ita.
3. The method according to claim 2, wherein a single training data set is applied to the dendritic neural network regression model to obtain a single first weight vector, wherein the feature quantity of the single first weight vector is the number of days multiplied by the sum of the parameters in the sample data set;
and calculating the length of a first weight vector of the dendritic neural network regression model according to the characteristic quantity and the branch number of the dendrobe layer.
4. The method of claim 3, wherein the loss function of the SFSMS algorithm is constructed according to the first weight vector and the actual value of the voltage transformer performance.
5. The method for analyzing the performance of the voltage transformer as claimed in claim 4, wherein a population set of the dendritic neural network regression model is constructed according to the first weight vector and an SFSMS algorithm, wherein the length of each first weight vector in the population set represents the weight and the threshold value of a training data set calculated through the dendritic neural network regression model.
6. The method for analyzing the performance of the voltage transformer as claimed in claim 5, wherein an iteration parameter t of an SFSMS algorithm is initialized to 1, parameters of a synapse layer and a cell layer of the dendritic neural network regression model are updated according to the SFSMS algorithm, and the loss weight value of the loss function corresponding to the first weight vector in the SFSMS algorithm is calculated;
updating the first weight vector in the population set according to the loss weight value to obtain a second weight vector of an individual;
and updating the second weight vector until the iteration parameter t is greater than the maximum iteration times to obtain a third weight vector representing the individual optimal weight result.
7. The method according to claim 1, wherein the predicted value is compared with the test data set to compare the prediction accuracy of the voltage transformer performance, wherein the element of the corresponding position of the predicted value after the third column vector of the test data set is flattened into one dimension is a performance parameter ratio error value for several days after prediction.
8. An analysis system for the performance of a voltage transformer, comprising:
the voltage transformer testing system comprises a data acquisition unit, a data acquisition unit and a data analysis unit, wherein the data acquisition unit is used for acquiring environmental parameters and operating parameters in an area where the voltage transformer is located, and constructing N sample data sets according to the environmental parameters and the operating parameters, wherein the N sample data sets comprise first N-1 training data sets and Nth testing data sets;
the model building unit is used for building a dendritic neural network regression model according to the SFSMS algorithm;
the first calculation unit is used for initializing the number i of training data sets to be 1, inputting the ith training data set into the dendritic neural network regression model for calculation to obtain a first weight vector corresponding to the ith training data set, and calculating a loss weight value corresponding to the SFSMS algorithm according to the first weight vector;
the second calculating unit is used for updating the first weight vector according to the loss weight value to obtain a second weight vector of an individual;
a judging unit, configured to set i to i +1, judge whether i is equal to N, if so, execute a third calculating unit, otherwise, repeatedly execute the first calculating unit and the second calculating unit, iteratively update the second weight vector until i is equal to N, and output a third weight vector representing an individual optimal weight result;
and the third calculating unit is used for calculating a predicted value of the Nth test data set according to the third weight vector and the (N-1) th training data set.
9. An electronic device, comprising: the system comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus; the memory for storing a computer program; the processor, configured to execute the program stored in the memory, and implement a method for analyzing the performance of the voltage transformer according to any one of claims 1 to 7.
10. A computer-readable storage medium storing computer-readable instructions that, when executed, perform the operations of a method of analyzing the performance of a voltage transformer of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210138807.6A CN114492199A (en) | 2022-02-15 | 2022-02-15 | Method and system for analyzing performance of voltage transformer, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210138807.6A CN114492199A (en) | 2022-02-15 | 2022-02-15 | Method and system for analyzing performance of voltage transformer, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114492199A true CN114492199A (en) | 2022-05-13 |
Family
ID=81481159
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210138807.6A Pending CN114492199A (en) | 2022-02-15 | 2022-02-15 | Method and system for analyzing performance of voltage transformer, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114492199A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056875A (en) * | 2023-10-10 | 2023-11-14 | 湖南华菱线缆股份有限公司 | Cable transmission performance analysis method and device based on test data |
-
2022
- 2022-02-15 CN CN202210138807.6A patent/CN114492199A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056875A (en) * | 2023-10-10 | 2023-11-14 | 湖南华菱线缆股份有限公司 | Cable transmission performance analysis method and device based on test data |
| CN117056875B (en) * | 2023-10-10 | 2024-01-02 | 湖南华菱线缆股份有限公司 | Cable transmission performance analysis method and device based on test data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111130839A (en) | Flow demand matrix prediction method and system | |
| CN106338708B (en) | Electric energy metering error analysis method combining deep learning and recurrent neural network | |
| CN113486078B (en) | Distributed power distribution network operation monitoring method and system | |
| CN114970362B (en) | Power grid load scheduling prediction method and system under multi-energy structure | |
| CN115201686B (en) | Lithium ion battery health state assessment method under incomplete charge and discharge data | |
| CN112529685A (en) | Loan user credit rating method and system based on BAS-FNN | |
| CN114781692A (en) | Short-term power load prediction method and device and electronic equipment | |
| CN114492199A (en) | Method and system for analyzing performance of voltage transformer, electronic equipment and medium | |
| CN115114843A (en) | Battery capacity prediction method, related device, equipment and storage medium | |
| CN108845546B (en) | Dynamic process monitoring method based on BP neural network autoregressive model | |
| CN115062528A (en) | Prediction method for industrial process time sequence data | |
| Zong et al. | Embedded software fault prediction based on back propagation neural network | |
| Yang et al. | A structure optimization algorithm of neural networks for large-scale data sets | |
| CN113093014A (en) | Online collaborative estimation method and system for SOH and SOC based on impedance parameters | |
| Song et al. | Capacity estimation method of lithium-ion batteries based on deep convolution neural network | |
| CN109187898B (en) | Soft measurement method and device for ammonia nitrogen content of water in aquaculture environment | |
| Hoomod et al. | Temperature Prediction Using Recurrent Neural Network for Internet of Things Room Controlling Application | |
| CN111160419B (en) | Deep learning-based electronic transformer data classification prediction method and device | |
| Ferrario et al. | Application of metamodel-based techniques for the efficient seismic analysis of structural systems | |
| Alrubaie | Cascade-Forward neural network for volterra integral equation solution | |
| CN108960406B (en) | MEMS gyroscope random error prediction method based on BFO wavelet neural network | |
| Shou-Ping et al. | An architecture of interval Elman network and its numerical analysis | |
| Draye et al. | Adaptative time constants improve the prediction capability of recurrent neural networks | |
| Roshchupkina et al. | ANFIS based approach for improved multisensors signal processing | |
| CN116359762B (en) | Battery state of charge estimation method based on deep learning and network compression |
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 |