CN108733921B - Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation - Google Patents
Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation Download PDFInfo
- Publication number
- CN108733921B CN108733921B CN201810481085.8A CN201810481085A CN108733921B CN 108733921 B CN108733921 B CN 108733921B CN 201810481085 A CN201810481085 A CN 201810481085A CN 108733921 B CN108733921 B CN 108733921B
- Authority
- CN
- China
- Prior art keywords
- granulation
- fuzzy information
- fuzzy
- data
- hot spot
- 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.)
- Active
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
-
- 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
- G06F2119/04—Ageing analysis or optimisation against ageing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation, which comprises the steps of determining the number of granulation time windows according to the fuzzy information granulation quality, and extracting effective information of original data by using a fuzzy information granulation technology; constructing a wavelet neural network prediction model based on the effective information, and optimizing the structural parameters of the wavelet neural network through a harmony search algorithm; and predicting the temperature fluctuation range of the hot spot of the transformer winding by using the model after structure and parameter adjustment. The method has high prediction accuracy and has certain guiding significance on the operation and maintenance of the transformer.
Description
Technical Field
The invention relates to a transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation.
Background
The power transformer is an important power transmission and transformation device, and has been an object of monitoring and protection. The loss of the proper insulation capacity is the main reason for the end of the life of most transformers, and the hot spot temperature of the transformer winding is one of the important factors influencing the insulation capacity. Winding hot spot temperature, defined as the temperature reached by the hottest areas of the transformer windings, has proven to be a key factor affecting the transformer load capacity and usable life. When the temperature of the hot spot of the transformer winding exceeds the allowable value, the insulation of the transformer can be damaged. Therefore, the hot spot temperature of the transformer is deeply researched, the abnormity of the hot spot temperature of the winding of the transformer is found in advance, and the method has very important significance for guiding the operation and maintenance of the transformer.
Experts and scholars at home and abroad carry out a great deal of research on prediction of the hot spot temperature of the transformer winding. Some of the existing methods provide a transformer top layer temperature prediction method based on a T-S fuzzy model, the model well completes the tracking of the top layer temperature and is compared with a recursion neural network and an IEEE guide rule, but the top layer temperature is not a real hot spot temperature, so that the model has certain limitation. Some documents consider 5 influence factors of ambient temperature, top layer oil temperature, bottom layer oil temperature, upper dead angle temperature and lower dead angle temperature, and a prediction model of the transformer winding hot spot temperature is established based on a generalized regression neural network. Some documents consider load transformation and top oil temperature, and a real-time estimation model of the transformer winding hot spot temperature is established based on a Kalman filtering algorithm. Some documents construct a transformer top-layer temperature prediction model based on a least square support vector machine. Some documents take the load current, the ambient temperature, the top layer oil temperature, the bottom layer oil temperature, the upper dead angle temperature and the lower dead angle temperature into consideration, and construct a hot spot temperature prediction model of the genetic optimization support vector machine.
The above researches well realize the tracking of the temperature of the transformer winding, but the models have relatively poor capability of predicting the abnormal temperature of the hot spot of the transformer winding in advance.
Disclosure of Invention
Aiming at the current situation that the time sequence prediction and the fluctuation range prediction research of the transformer winding hot spot temperature are relatively incomplete, a transformer winding hot spot temperature fluctuation range combined prediction model of a Wavelet Neural Network (WNN) is optimized based on a Harmony Search algorithm (HS), and the capability of predicting the transformer winding hot spot temperature abnormity can be better predicted.
In order to achieve the purpose, the invention adopts the following technical scheme:
a transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation comprises the following steps:
determining the number of granulation time windows according to the fuzzy information granulation quality, and extracting effective information of the original data by using a fuzzy information granulation technology;
constructing a wavelet neural network prediction model based on the effective information, and optimizing the structural parameters of the wavelet neural network through a harmony search algorithm;
and predicting the temperature fluctuation range of the hot spot of the transformer winding by using the model after structure and parameter adjustment.
In more detail, the method comprises the following steps:
updating historical data, carrying out fuzzy information granulation on historical data sample data to obtain an average value, a lower bound and an upper bound, calculating fuzzy information granulation quality factors, and selecting an optimal granulation result;
screening granulated historical data, preliminarily establishing wavelet neural network prediction models with different structures by using training set samples, judging the number of nodes of an input layer and the number of nodes of a hidden layer through the models, and updating at regular time;
determining an input layer, a hidden layer and an output layer, namely determining structural parameters needing to be optimized, encoding an optimized object, defining a fitness value function, and screening the structural parameters of the wavelet neural network through harmony search;
predicting the average value, the lower bound and the upper bound respectively by using the screened wavelet neural network to obtain the temperature fluctuation range of the hot spot of the transformer winding;
and evaluating the model prediction performance by adopting three indexes of mean square error, mean absolute error and correlation coefficient.
Further, the process of extracting the effective information of the original data by using the fuzzy information granulation technology specifically comprises the following steps:
decomposing the transformer winding hot spot temperature time sequence into a plurality of small subsequences according to the requirement, and fuzzifying each subsequence as an operation time window;
establishing a triangular fuzzy particle on each subsequence;
and sequencing the time sequence from small to large, and solving the average value, the lower bound and the upper bound of each triangular fuzzy particle.
Further, in the process of extracting effective information of the original data by using a fuzzy information granulation technology, the number of granulation time windows is determined by using fuzzy information granulation quality factors, wherein the quality factors are log values of data compression ratios of fuzzy information granulation and loss ratios of data after fuzzy information granulation.
Furthermore, in the wavelet neural network prediction model constructed based on the effective information, a Morlet basis wavelet function is adopted, and the historical data of the transformer winding hot spot temperature after fuzzy granulation is used as the input data of the input layer node of the wavelet neural network.
Further, in searching the optimal wavelet neural network structure parameters by using a harmony search algorithm, the prediction error of the wavelet neural network model is used as a target function, namely, the wavelet neural network is called to predict, the obtained prediction result is used as the harmony fitness value, and the input weight, the output weight, the scaling factor and the translation factor of the wavelet neural network to be optimized are used as the position vector of each harmony.
Further, the optimizing step includes:
(1) initializing the capacity of a harmony memory bank, the value probability of the memory bank, the tone fine-tuning probability, the tone fine-tuning broadband and the maximum iteration number;
(2) generating a plurality of harmony to form an initial harmony library, calling a wavelet neural network to predict, and calculating prediction errors to obtain fitness values of all bodies in an original memory library;
(3) randomly generating a random number of 0-1, and randomly selecting an individual from a memory bank if the random number is less than the value probability of the memory bank;
(4) fine-tuning the selected individuals; if the random number is not smaller than the value probability of the memory bank, a new solution is regenerated in the value range of the variable;
(5) updating the harmony memory base, calculating the fitness value of the generated new solution, and updating the memory base;
(6) and (4) judging whether the algorithm is terminated, if the maximum iteration number is met, terminating the algorithm to obtain the optimal wavelet neural network structure parameters, and if not, returning to the step (3) to continue execution.
Further, in the step (3), the process of randomly selecting an individual from the memory pool includes: if the random number is smaller than the value probability of the memory bank, the new individual is one of harmony, otherwise, the new individual is one of harmony vectors.
Further, in the step (4), if the random number is smaller than the value probability of the memory bank, the new solution is:
rand2 is a random number between 0 and1, k is the number of possible values of a discrete variable, and PAR is the pitch trimming probability.
Further, in the step (5), a fitness value of the new solution is calculated, and the memory bank is updated according to the fitness value:
xworstthe function expression f () here is the objective function for the worst solution in the solution set.
Compared with the prior art, the invention has the beneficial effects that:
(1) the information granulation technology can extract useful information from the original data as required, the complexity of target data is reduced, and the wavelet neural network has strong nonlinear mapping capability. The invention combines the two methods to effectively realize the prediction of the temperature fluctuation range of the hot spot of the transformer winding.
(2) The proposed concept of fuzzy information granulation quality optimizes the information granulation process; structural parameters of the wavelet neural network are screened by using a harmony search algorithm, prediction accuracy is improved, the structural parameters are compared with various prediction models, the superiority of the model provided by the invention is verified, and the feasibility of the FIG-HS-WNN model for predicting the hot spot temperature of the transformer winding is demonstrated.
(3) The method has high prediction accuracy and has certain guiding significance on the operation and maintenance of the transformer. The combined model can be used for predicting the temperature of the hot spot of the transformer winding and can provide ideas for prediction modeling in other fields.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
FIG. 1 is a diagram of the topology of a wavelet neural network of the present invention;
FIG. 2 is a flow diagram of the combined predictive model of the present invention;
the specific implementation mode is as follows:
the invention is further described with reference to the following figures and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
In the present invention, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only terms of relationships determined for convenience of describing structural relationships of the parts or elements of the present invention, and are not intended to refer to any parts or elements of the present invention, and are not to be construed as limiting the present invention.
In the present invention, terms such as "fixedly connected", "connected", and the like are to be understood in a broad sense, and mean either a fixed connection or an integrally connected or detachable connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be determined according to specific situations by persons skilled in the relevant scientific or technical field, and are not to be construed as limiting the present invention.
Aiming at the current situation that the research on the time sequence prediction and fluctuation range prediction of the hot spot temperature of the transformer winding is relatively incomplete, a combined prediction model of the temperature fluctuation range of the hot spot of the transformer winding based on fuzzy information granulation and Harmony Search algorithm (HS) for optimizing Wavelet Neural Network (WNN) is provided.
Firstly, determining the number of granulation time windows according to the fuzzy information granulation quality, and further extracting effective information of original data by using a fuzzy information granulation technology;
then, a wavelet neural network prediction model is constructed based on the effective information, and structural parameters of the wavelet neural network are optimized through a harmony search algorithm;
and finally, predicting the temperature fluctuation range of the hot spot of the transformer winding by using the model with the adjusted structure and parameters. The experiment and data analysis result shows that the model can effectively predict the temperature fluctuation range of the winding hot spot of the transformer and has certain guiding significance on the operation and maintenance of the transformer.
Fuzzy information granulation
1.1 fuzzy information granulation rationale
Information granulation was first proposed by professor Zadeh, which is a process of breaking down a large amount of complex information into a plurality of parts according to a certain rule, each part having a low computational complexity. The method realizes the compression of the original information while extracting the effective data of the original information. The information granulation mainly comprises three types, namely a fuzzy set theory, a rough set theory and an entropy space theory, and the fuzzy granulation process of the transformer winding hot spot temperature time sequence is researched by adopting a model based on the fuzzy set theory. Fuzzy information granulation is divided into two processes: (1) dividing a granulation time window; (2) and fuzzifying the data in each time window according to a certain rule.
The granulation time window division is to decompose the transformer winding hot spot temperature time sequence into a plurality of small subsequences according to requirements, and each subsequence is used as an operation time window. The information fuzzification is an effective information extraction process, and a reasonable fuzzy set is established for data in each time window according to a certain rule.
Consider the case of blurring a single time window, i.e., regarding a transformer winding hot spot temperature sequence T ═ T1, T2, …, tn ] as a time window. The goal of the blurring is to create a blurring particle P on T, i.e. to find a certain function a such that:
P＝A(t),t∈T (1)
the fuzzy particle P mainly has basic forms of triangle, trapezoid, asymmetric high-gaussian and the like, and because the invention predicts the fluctuation range of the hot spot temperature of the transformer winding, namely the maximum value and the minimum value of a window need to be known, the triangular fuzzy particle is adopted, and the membership function of the triangular fuzzy particle is shown as the formula (2):
in the fuzzification process, namely, the values of LOW, R and UP are determined, the invention adopts a W.Pedracz algorithm:
(1) and (5) calculating an average value R. And (4) sequencing the time series T from small to large, and if the time series T is not sequenced and is still T, then R is mean (T).
(2) Solving a lower bound LOW:
(3) and (4) solving an upper bound UP:
(4) giving a blur particle P:
P＝(LOW,R,UP) (5)
the values of the parameters LOW, R, UP can be determined by the above process, where the parameter LOW represents the minimum value of the data change in each time window, the parameter R represents the average level of the data change in each time window, and the parameter UP represents the maximum value of the data change in each time window.
1.2 fuzzy information graining quality
When fuzzy information granulation is applied, the number of the granulation time windows needs to be given artificially, and the granulation time windows inevitably have personal subjective colors, so that information redundancy or loss is easily caused. In order to solve the problem, the invention provides a fuzzy information granulation quality factor, and the number of granulation time windows is determined from a more scientific angle.
Setting a winding hot spot temperature time sequence T ═ T1, T2, …, tn ], after fuzzy information granulation LOW ═ a1, a2, …, al ], R ═ m1, m2, …, ml ], UP ═ b1, b2, …, bl.
The data compression ratio for defining fuzzy information granulation is as follows:
CR＝n/(3l) (6)
in the formula (6), n is the length of the original sequence T, l is the length of LOW, R and UP, and the larger CR is, the higher the data compression degree is.
The entropy of information can represent to some extent how much information a group of data carries. For this reason, information entropy is introduced, and the information quantity carried by the data before and after the granulation is described. Solving the information entropy of T, LOW, R and UP, and setting HT, HLOW, HR and HUP respectively, so that the loss of the data after fuzzy information granulation is defined as:
the larger the S, the more serious the data loss after the fuzzy information is granulated.
The fuzzy information granulation figure of merit is defined as:
the higher the quality factor is, the better the fuzzy information granulation result is, i.e. a balance point is found between the data compression ratio and the data loss amount, and it is desirable that the result after fuzzy information granulation has a certain compression ratio while the data loss amount is small.
Optimization of wavelet neural network by using dyadic search algorithm
2.1 wavelet neural network
The wavelet neural network is a product of combining wavelet analysis and the neural network, inherits the time-frequency local property of wavelet transformation, and has the self-learning and nonlinear mapping capabilities of the neural network. The wavelet neural network is divided according to the combination mode of the two, the wavelet neural network comprises a loose wavelet neural network and a compact wavelet neural network, the loose wavelet neural network presents wavelet preprocessing and the neural network in a series connection mode, and the compact wavelet neural network completes the design of hidden layer nodes of the neural network by utilizing wavelet transformation. The invention develops around a compact wavelet neural network.
Due to the nonlinear characteristic of the temperature of the hot spot of the transformer winding, the characteristic extraction and nonlinear fitting capability of the wavelet neural network is suitable. The invention adopts a single-layer output wavelet neural network, and the network topology structure of the single-layer output wavelet neural network is shown in figure 1.
In fig. 1, { x1, …, xi, …, xn } represents input data of input layer nodes of the wavelet neural network, i.e. history data of transformer winding hot spot temperature after fuzzy granulation, y represents node prediction output data of output layer of the wavelet neural network, wij represents weight coefficient from input layer to hidden layer, and wj represents weight coefficient from hidden layer to output layer. Ψ is the excitation function of the hidden layer node, here using the Morlet basis wavelet function.
Assuming that the ith node input is xi (i ═ 1,2, …, n), the input value of the jth hidden layer node is:
after data enters a hidden layer node, the data is subjected to expansion and translation transformation, and then the input value of the basic wavelet function is as follows:
(10) in the formula, aj is a scale factor, and bj is a translation factor. And the Morlet basic wavelet function is shown as a formula (11), and the hidden layer node output result is shown as a formula (12).
The output node neurons of the wavelet neural network adopt Sigmoid functions, that is, f (x) is 1/(1+ e-x), and the final model prediction result can be expressed as expression (13).
The prediction error of the defined model is:
in equation (14), T is the number of training samples, YP represents measured data of the P-th sample, and YP represents predicted data of the P-th sample.
The weights wij and wj of the WNN, the scaling factor aj and the translation factor bj have a large influence on the prediction error J, so that reasonable model parameters need to be selected. The traditional wavelet neural network selects parameters by a gradient descent method of additional momentum to minimize the formula (14), but the method has the problems of low convergence speed and incapability of effectively searching global optimum. In order to effectively solve the problem, the invention researches a harmony search algorithm.
2.2 HS optimization of WNN structural parameters
The harmony search algorithm simulates the principle of musical performance, has strong global search capability, avoids the problem of complex parameter setting, and has more excellent performance than a genetic algorithm and a simulated annealing algorithm in a plurality of optimization problems. The method optimizes the structural parameters of the wavelet neural network by using the algorithm, and improves the accuracy of winding hot spot temperature prediction. The prior art has the literature describing the principles of the harmony search algorithm in detail, and the present invention is not repeated.
The optimal wavelet neural network structure parameters are searched by utilizing a harmony search algorithm, and the problems of establishment of fitness value functions and harmony coding need to be considered emphatically. In the invention, the prediction error of the wavelet neural network model is used as a target function during design, namely, the wavelet neural network is called for prediction, and the result obtained by the formula (14) is used as the harmonic fitness value. And taking the input weight wij, the output weight wj, the scaling factor aj and the translation factor bj which need to be optimized as the position vector of each harmony, as shown in a formula (15).
X(HMS)＝[wij,aj,bj,wj] (15)
In the formula, HMS is the number of harmony, i is the number of nodes in the input layer, j is the number of nodes in the hidden layer, and the harmony dimension D is (i +3) × j.
The optimization steps are as follows:
(1) and initializing parameters. The parameters to be set are: and the volume of a sound memory library HM, the value probability HMCR of the memory library, the tone fine tuning probability PAR, the tone fine tuning broadband bw and the maximum iteration time Tmax.
(2) Initialization and sound library. HMS harmonics are generated, constituting an initial harmonic library. And calling a wavelet neural network to predict, and calculating a prediction error to obtain each body fitness value in an original memory library (HM).
(3) A new harmony sound is generated. Randomly generating a random number rand1 of 0-1, and if rand1< HMCR is satisfied, randomly selecting an individual from a memory bank (HM) according to formula (16).
Fine-tuning the selected individuals according to the formula (17); if rand1< HMCR is not satisfied, a new solution is regenerated within the value range of the variable.
In the formula (17), rand2 is a random number between 0 and1, and k is the number of possible values of the discrete variable.
(4) Update and sound memory bank. The fitness value of the new solution generated in step 3 is calculated and the memory (HM) is updated according to the following formula.
(5) It is determined whether the algorithm is terminated. If the maximum iteration times are met, the algorithm is terminated, and the optimal wavelet neural network structure parameters are obtained. Otherwise, the algorithm goes to step 3 to continue execution.
Three-hot-spot temperature fluctuation range prediction modeling
On the basis of the work, a transformer winding hot spot temperature fluctuation range combined prediction model based on fuzzy information granulation and harmony search optimization wavelet neural network (FIG-HS-WNN) is constructed. The overall flow of the model is shown in fig. 2.
The modeling steps are briefly described as follows:
(1) updating historical data, carrying out fuzzy information granulation on the sample data according to the formulas (2) to (5) to obtain LOW, R and UP, calculating a fuzzy information granulation quality factor Q according to the formulas (6) to (8), and selecting an optimal granulation result.
(2) Screening the granulated historical data, preliminarily establishing wavelet neural network prediction models with different structures by using training set samples, and judging the number of nodes of an input layer (namely the number of days for inputting the historical data) and the number of nodes of a hidden layer by using the models. It should be noted that the structure of the wavelet neural network may change for different historical data sets, i.e. the wavelet neural network needs to be readjusted after data is updated every day.
(3) Determining an input layer, a hidden layer and an output layer, namely determining structural parameters needing to be optimized, encoding an optimized object according to an equation (15), defining an equation (14) as a fitness value function, and screening the structural parameters of the wavelet neural network through harmony search.
(4) And predicting LOW, R and UP respectively by using the designed wavelet neural network to obtain the temperature fluctuation range of the hot point of the transformer winding.
(5) And evaluating the prediction performance of the model by using three indexes, namely Mean Square Error (MSE), Mean Absolute Error (MAE) and correlation coefficient (r).
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (9)
1. A transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation is characterized by comprising the following steps: the method comprises the following steps:
determining the number of granulation time windows according to the fuzzy information granulation quality, and extracting effective information of the original data by using a fuzzy information granulation technology;
constructing a wavelet neural network prediction model based on the effective information, and optimizing the structural parameters of the wavelet neural network through a harmony search algorithm;
predicting the temperature fluctuation range of the hot point of the transformer winding by using the model with the adjusted structure and parameters;
the method for determining the number of the granulation time windows according to the fuzzy information granulation quality and extracting the effective information of the original data by using the fuzzy information granulation technology comprises the following steps:
setting a winding hot spot temperature time sequence T ═ T1, T2, …, tn ], after fuzzy information granulation LOW ═ a1, a2, …, al ], R ═ m1, m2, …, ml ], UP ═ b1, b2, …, bl;
the data compression ratio for defining fuzzy information granulation is as follows:
CR＝n/(3l) (6)
in formula (6), n is the length of original sequence T, l is the length of LOW, R, UP, CRThe larger the data compression degree is, the higher the data compression degree is;
introducing information entropy, describing information quantity carried by data before and after granulation, and solving the information entropy of T, LOW, R and UP, which are respectively set as HT、HLOW、HR、HUPThen, the loss amount of the data after the fuzzy information is granulated is defined as:
the larger the S is, the more serious the data loss is after the fuzzy information is granulated;
the fuzzy information granulation figure of merit is defined as:
the higher the quality factor is, the better the fuzzy information granulation result is, i.e. a balance point is found between the data compression ratio and the data loss amount, and it is desirable that the result after fuzzy information granulation has a certain compression ratio while the data loss amount is small.
2. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 1, wherein: the process of extracting effective information of the original data by using the fuzzy information granulation technology specifically comprises the following steps:
decomposing the transformer winding hot spot temperature time sequence into a plurality of small subsequences according to the requirement, and fuzzifying each subsequence as an operation time window;
establishing a triangular fuzzy particle on each subsequence;
and sequencing the time sequence from small to large, and solving the average value, the lower bound and the upper bound of each triangular fuzzy particle.
3. A transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation is characterized by comprising the following steps: the method comprises the following steps:
updating historical data, carrying out fuzzy information granulation on historical data sample data to obtain an average value, a lower bound and an upper bound, calculating fuzzy information granulation quality factors, and selecting an optimal granulation result;
screening granulated historical data, preliminarily establishing wavelet neural network prediction models with different structures by using training set samples, judging the number of nodes of an input layer and the number of nodes of a hidden layer through the models, and updating at regular time;
determining an input layer, a hidden layer and an output layer, namely determining structural parameters needing to be optimized, encoding an optimized object, defining a fitness value function, and screening the structural parameters of the wavelet neural network through harmony search;
predicting the average value, the lower bound and the upper bound respectively by using the screened wavelet neural network to obtain the temperature fluctuation range of the hot spot of the transformer winding;
evaluating the prediction performance of the model by three indexes of mean square error, mean absolute error and correlation coefficient;
the historical data is updated, fuzzy information granulation is carried out on historical data sample data to obtain an average value, a lower bound and an upper bound, fuzzy information granulation quality factors are calculated, and an optimal granulation result is selected, wherein the fuzzy information granulation quality factors comprise:
setting a winding hot spot temperature time sequence T ═ T1, T2, …, tn ], after fuzzy information granulation LOW ═ a1, a2, …, al ], R ═ m1, m2, …, ml ], UP ═ b1, b2, …, bl;
the data compression ratio for defining fuzzy information granulation is as follows:
CR＝n/(3l) (6)
in formula (6), n is the length of original sequence T, l is the length of LOW, R, UP, CRThe larger the data compression degree is, the higher the data compression degree is;
introducing information entropy, describing information quantity carried by data before and after granulation, and solving the information entropy of T, LOW, R and UP, which are respectively set as HT、HLOW、HR、HUPThen, the loss amount of the data after the fuzzy information is granulated is defined as:
the larger the S is, the more serious the data loss is after the fuzzy information is granulated;
the fuzzy information granulation figure of merit is defined as:
the higher the quality factor is, the better the fuzzy information granulation result is, i.e. a balance point is found between the data compression ratio and the data loss amount, and it is desirable that the result after fuzzy information granulation has a certain compression ratio while the data loss amount is small.
4. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 1 or 3, characterized by: in the wavelet neural network prediction model constructed based on the effective information, a Morlet basis wavelet function is adopted, and the historical data of the transformer winding hot spot temperature after fuzzy granulation is used as the input data of the input layer nodes of the wavelet neural network.
5. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 1 or 3, characterized by: in the searching of the optimal wavelet neural network structure parameters by utilizing the harmony search algorithm, the prediction error of the wavelet neural network model is used as a target function, namely, the wavelet neural network is called to predict, the obtained prediction result is used as the harmony fitness value, and the input weight, the output weight, the scaling factor and the translation factor of the wavelet neural network to be optimized are used as the position vector of each harmony.
6. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 5, wherein: the optimization steps comprise:
(1) initializing the capacity of a harmony memory bank, the value probability of the memory bank, the tone fine-tuning probability, the tone fine-tuning broadband and the maximum iteration number;
(2) generating a plurality of harmony to form an initial harmony library, calling a wavelet neural network to predict, and calculating prediction errors to obtain fitness values of all bodies in an original memory library;
(3) randomly generating a random number of 0-1, and randomly selecting an individual from a memory bank if the random number is less than the value probability of the memory bank;
(4) fine-tuning the selected individuals; if the random number is not smaller than the value probability of the memory bank, a new solution is regenerated in the value range of the variable;
(5) updating the harmony memory base, calculating the fitness value of the generated new solution, and updating the memory base;
and (4) judging whether the algorithm is terminated, if the maximum iteration number is met, terminating the algorithm to obtain the optimal wavelet neural network structure parameters, and if not, returning to the step (3) to continue execution.
7. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 6, wherein: in the step (3), the process of randomly selecting an individual from the memory bank includes: if the random number is smaller than the value probability of the memory bank, the new individual is one of harmony, otherwise, the new individual is one of harmony vectors.
8. The transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation as claimed in claim 6, wherein: in the step (4), if the random number is smaller than the value probability of the memory bank, the new solution is:
rand2 is a random number between 0 and1, k is the number of possible values of a discrete variable, and PAR is the pitch trimming probability.
9. The method for predicting the temperature fluctuation range of the hot spot of the transformer winding based on fuzzy information granulation as claimed in claim 8, wherein: in the step (5), calculating the fitness value of the new solution, and updating the memory base according to the fitness value:
xworstthe worst solution in the solution set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810481085.8A CN108733921B (en) | 2018-05-18 | 2018-05-18 | Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810481085.8A CN108733921B (en) | 2018-05-18 | 2018-05-18 | Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108733921A CN108733921A (en) | 2018-11-02 |
| CN108733921B true CN108733921B (en) | 2021-07-16 |
Family
ID=63938458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810481085.8A Active CN108733921B (en) | 2018-05-18 | 2018-05-18 | Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108733921B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109375504B (en) * | 2018-11-06 | 2022-04-19 | 江西北斗变电科技有限公司 | Winding insulation bubble generation control method based on fuzzy neural network PID control |
| CN109901030B (en) * | 2019-02-25 | 2021-10-22 | 国网山东省电力公司济南供电公司 | Reactor turn-to-turn insulation state monitoring method, system and application |
| CN110244225A (en) * | 2019-06-27 | 2019-09-17 | 清华大学深圳研究生院 | A method of hot face temperature when prediction power battery electric discharge |
| CN116432406B (en) * | 2023-03-09 | 2024-02-02 | 广东电网有限责任公司佛山供电局 | Method and device for calculating hot spot temperature of working winding of oil immersed transformer |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6609079B1 (en) * | 1998-05-14 | 2003-08-19 | Va Tech Elin Transformatoren Gmbh | Method and arrangement for ascertaining state variables |
| CN102915355A (en) * | 2012-10-11 | 2013-02-06 | 李英明 | Multiprocessor task scheduling method based on harmony search and simulated annealing |
| CN103150610A (en) * | 2013-02-28 | 2013-06-12 | 哈尔滨工业大学 | Fuzzy information granulation and support vector machine-based heating load prediction method |
| CN105550472A (en) * | 2016-01-20 | 2016-05-04 | 国网上海市电力公司 | Prediction method of transformer winding hot-spot temperature based on neural network |
| CN105590023A (en) * | 2015-12-08 | 2016-05-18 | 三峡大学 | Fuzzy granulation prediction method of performance degradation of rolling bearing on the basis of information entropy |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6445372B2 (en) * | 2015-03-31 | 2018-12-26 | 日本碍子株式会社 | Piezoelectric / electrostrictive material, piezoelectric / electrostrictive body, and resonance driving device |
| CN106073702B (en) * | 2016-05-27 | 2019-05-28 | 燕山大学 | Based on the more time-frequency scale diencephalon myoelectricity coupling analytical methods of small echo-transfer entropy |
| CN106527141B (en) * | 2016-12-05 | 2019-09-20 | 清华大学 | Air/Fuel Ratio in Glass Furnace method of adjustment based on variable universe fuzzy rule iterative learning |
| CN107784655A (en) * | 2016-12-28 | 2018-03-09 | 中国测绘科学研究院 | A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold |
| CN107368926B (en) * | 2017-07-28 | 2018-07-10 | 中南大学 | A kind of more natural parameter sensing method for amalgamation processing of intelligent environment carrying robot identification floor |
| CN107506871A (en) * | 2017-09-08 | 2017-12-22 | 广东工业大学 | A kind of method and system of interval prediction |
-
2018
- 2018-05-18 CN CN201810481085.8A patent/CN108733921B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6609079B1 (en) * | 1998-05-14 | 2003-08-19 | Va Tech Elin Transformatoren Gmbh | Method and arrangement for ascertaining state variables |
| CN102915355A (en) * | 2012-10-11 | 2013-02-06 | 李英明 | Multiprocessor task scheduling method based on harmony search and simulated annealing |
| CN103150610A (en) * | 2013-02-28 | 2013-06-12 | 哈尔滨工业大学 | Fuzzy information granulation and support vector machine-based heating load prediction method |
| CN105590023A (en) * | 2015-12-08 | 2016-05-18 | 三峡大学 | Fuzzy granulation prediction method of performance degradation of rolling bearing on the basis of information entropy |
| CN105550472A (en) * | 2016-01-20 | 2016-05-04 | 国网上海市电力公司 | Prediction method of transformer winding hot-spot temperature based on neural network |
Non-Patent Citations (2)
| Title |
|---|
| A New Prediction Model for Transformer Winding Hotspot Temperature Fluctuation Based on Fuzzy Information Granulation and an Optimized Wavelet Neural Network;Li Zhang 等;《energies》;20171201;第10卷(第12期);1-13 * |
| 基于改进变分模态分解的有载分接开关机械状态监测;王冠 等;《湖南大学学报(自然科学版)》;20171031;第44卷(第10期);75-83 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108733921A (en) | 2018-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108733921B (en) | Transformer winding hot spot temperature fluctuation range prediction method based on fuzzy information granulation | |
| Cordón et al. | Hybridizing genetic algorithms with sharing scheme and evolution strategies for designing approximate fuzzy rule-based systems | |
| CN106529818B (en) | Water quality assessment Forecasting Methodology based on Fuzzy Wavelet Network | |
| CN108876044B (en) | Online content popularity prediction method based on knowledge-enhanced neural network | |
| CN110851566A (en) | Improved differentiable network structure searching method | |
| CN109978283A (en) | Photovoltaic power generation power prediction method based on branch evolution neural network | |
| CN111832825A (en) | Wind power prediction method and system integrating long-term and short-term memory network and extreme learning machine | |
| CN114998525A (en) | Action identification method based on dynamic local-global graph convolutional neural network | |
| CN111355633A (en) | Mobile phone internet traffic prediction method in competition venue based on PSO-DELM algorithm | |
| Cai et al. | On-device image classification with proxyless neural architecture search and quantization-aware fine-tuning | |
| Huang et al. | Ponas: Progressive one-shot neural architecture search for very efficient deployment | |
| CN112036651A (en) | Electricity price prediction method based on quantum immune optimization BP neural network algorithm | |
| Lan et al. | A spectrum prediction approach based on neural networks optimized by genetic algorithm in cognitive radio networks | |
| Cao et al. | An enhanced whale optimization algorithm with improved dynamic opposite learning and adaptive inertia weight strategy | |
| CN115796358A (en) | Carbon emission prediction method and terminal | |
| CN116596109A (en) | Traffic flow prediction model based on gating time convolution network | |
| CN112183721A (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
| CN116821452A (en) | Graph node classification model training method and graph node classification method | |
| Lavaei et al. | Dynamic analysis of structures using neural networks | |
| Shubha et al. | A diverse noise-resilient DNN ensemble model on edge devices for time-series data | |
| Wang et al. | Elongation prediction of steel-strips in annealing furnace with deep learning via improved incremental extreme learning machine | |
| Lu et al. | Laplacian deep echo state network optimized by genetic algorithm | |
| CN114625886A (en) | Entity query method and system based on knowledge graph small sample relation learning model | |
| Wang et al. | Interval-valued financial time series prediction based on improved Elman neural network | |
| Mirzaei et al. | Optimal matching by the transiently chaotic neural network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |