CN115796843A - Operation and maintenance strategy generation method of power transformer and related device - Google Patents
Operation and maintenance strategy generation method of power transformer and related device Download PDFInfo
- Publication number
- CN115796843A CN115796843A CN202211591750.1A CN202211591750A CN115796843A CN 115796843 A CN115796843 A CN 115796843A CN 202211591750 A CN202211591750 A CN 202211591750A CN 115796843 A CN115796843 A CN 115796843A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- network model
- equipment
- power transformer
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses an operation and maintenance strategy generation method and a related device of a power transformer, wherein the method comprises the following steps: respectively acquiring state quantity data and transformer basic information of different parts of a power transformer; extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence; training a preset convolutional neural network model by adopting a two-dimensional image training data set to obtain an optimized convolutional neural network model, and optimally training the parameters of the optimized model of the preset convolutional neural network model according to an artificial fish swarm-firefly algorithm to obtain the optimized convolutional neural network model; and inputting the currently acquired real-time comprehensive data information into an optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy. The method and the device can solve the technical problems that in the prior art, operation and maintenance strategies cannot be designed aiming at the component health state of the power transformer, the operation and maintenance strategies obtained actually lack pertinence, and the effect is poor.
Description
Technical Field
The present application relates to the field of power transformers, and in particular, to a method and an apparatus for generating an operation and maintenance policy of a power transformer.
Background
The electric power industry is a basic energy industry related to the national civilization and has important significance for the healthy development of national economy and the improvement of the living standard of people. The power transformer is a junction for electric energy transmission and conversion in the power system, and the safety and reliability of the transformer are the basis for the safe and stable operation of the whole power system. For a long time, domestic and foreign power grid companies all put a great deal of effort in formulating operation and maintenance strategies for power transformers.
The existing differentiated operation and maintenance scheme mainly comprises the steps that an operation and maintenance data acquisition specially-assigned person can analyze data after collecting the data, the health state of equipment is evaluated after defective components are classified, then a control level is determined based on the health state and the importance of the equipment, and finally an operation and maintenance strategy is formulated and executed based on the control level. However, an operation and maintenance strategy suitable for the detailed health state of each component of the power transformer cannot be formulated, so that the operation and maintenance strategy of the power transformer lacks pertinence, and the problems of excessive operation and maintenance of the power transformer, insufficient operation and maintenance of the power transformer, waste of operation and maintenance resources, low operation and maintenance performance and the like are caused.
Disclosure of Invention
The application provides an operation and maintenance strategy generation method and a related device for a power transformer, which are used for solving the technical problems that the operation and maintenance strategy obtained actually is lack of pertinence and poor in effect due to the fact that the operation and maintenance strategy design cannot be carried out according to the component health state of the power transformer in the prior art.
In view of this, a first aspect of the present application provides a method for generating an operation and maintenance policy of a power transformer, including:
respectively acquiring state quantity data and transformer basic information of different parts of a power transformer;
extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and a time sequence;
training a preset convolutional neural network model by adopting the two-dimensional image training data set to obtain an optimized convolutional neural network model, wherein the optimized model parameters of the preset convolutional neural network model are obtained by optimized training according to an artificial fish swarm-firefly algorithm;
and inputting the currently acquired real-time comprehensive data information into the optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
Preferably, the extracting the comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence includes:
calculating a withholding value of each component according to the state quantity data, and extracting the withholding value of each component based on preset weight;
performing state evaluation on each part according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating the fault probability of the equipment by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on the equipment price, the load importance and the equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
Preferably, the converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set, before further comprising:
and carrying out exception rejection, normalization and one-hot coding processing on the comprehensive data information to realize data information standardization.
Preferably, the training of the preset convolutional neural network model by using the two-dimensional image training data set to obtain an optimized convolutional neural network model, where the optimized model parameters of the preset convolutional neural network model are obtained by performing optimized training according to an artificial fish swarm-firefly algorithm, and the method also includes:
constructing an initial convolutional neural network model based on an attention mechanism;
performing individual sharing-based joint optimization training on initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and adjusting the initial convolutional neural network model through the optimization model parameters to obtain a preset convolutional neural network model.
The second aspect of the present application provides an operation and maintenance strategy generation apparatus for a power transformer, including:
the information acquisition unit is used for respectively acquiring state quantity data and transformer basic information of different parts of the power transformer;
the information processing unit is used for extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence;
the model training unit is used for training a preset convolutional neural network model by adopting the two-dimensional image training data set to obtain an optimized convolutional neural network model, and the optimized model parameters of the preset convolutional neural network model are obtained by optimized training according to an artificial fish swarm-firefly algorithm;
and the strategy generating unit is used for inputting the currently acquired real-time comprehensive data information into the optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
Preferably, the information processing unit is specifically configured to:
calculating a withholding value of each component according to the state quantity data, and extracting the withholding value of each component based on preset weight;
performing state evaluation on each part according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating the fault probability of the equipment by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on equipment price, load importance and equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
Preferably, the method further comprises the following steps:
and the standardization unit is used for carrying out exception removal, normalization and one-hot coding processing on the comprehensive data information to realize data information standardization.
Preferably, the method further comprises the following steps:
the model building unit is used for building an initial convolutional neural network model based on an attention mechanism;
the parameter optimization unit is used for performing joint optimization training based on individual sharing on the initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and the model optimization unit is used for adjusting the initial convolutional neural network model through the optimization model parameters to obtain a preset convolutional neural network model.
A third aspect of the present application provides an operation and maintenance strategy generation device for a power transformer, where the device includes a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the operation and maintenance strategy generation method for the power transformer according to the first aspect according to instructions in the program code.
A fourth aspect of the present application provides a computer-readable storage medium for storing program codes, where the program codes are used to execute the operation and maintenance strategy generation method for the power transformer of the first aspect.
According to the technical scheme, the embodiment of the application has the following advantages:
the application provides an operation and maintenance strategy generation method of a power transformer, which comprises the following steps: respectively acquiring state quantity data and transformer basic information of different parts of a power transformer; extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence; training a preset convolutional neural network model by adopting a two-dimensional image training data set to obtain an optimized convolutional neural network model, and optimally training the parameters of the optimized model of the preset convolutional neural network model according to an artificial fish swarm-firefly algorithm to obtain the optimized convolutional neural network model; and inputting the currently acquired real-time comprehensive data information into an optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
According to the operation and maintenance strategy generation method of the power transformer, operation and maintenance data analysis is carried out by acquiring state quantity data of different parts of the power transformer and transformer basic information, and then an optimized neural network model is trained according to comprehensive data information formed by various parts and equipment data, so that the model has stronger expression capacity in operation and maintenance strategy generation, and the accuracy of the generated optimized operation and maintenance strategy can be ensured; the model parameters are optimally trained by adopting a specific algorithm, so that the learning performance of the model can be improved, and the reliability of the model is further ensured. Therefore, the method and the device can solve the technical problems that operation and maintenance strategies which are obtained actually lack pertinence and are poor in effect due to the fact that operation and maintenance strategy design cannot be carried out on the component health state of the power transformer in the prior art.
Drawings
Fig. 1 is a schematic flowchart of an operation and maintenance strategy generation method for a power transformer according to an embodiment of the present disclosure;
fig. 2 is a schematic structural diagram of an operation and maintenance strategy generation apparatus for a power transformer according to an embodiment of the present application;
FIG. 3 is an exemplary diagram of a component state quantity deduction standard provided by an embodiment of the present application;
FIG. 4 is a schematic diagram of a process flow of comprehensive data information provided in an embodiment of the present application;
FIG. 5 is a schematic view of an attention mechanism according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of an initial convolutional neural network model structure provided in an embodiment of the present application;
fig. 7 is a schematic diagram of a preset convolutional neural network model data processing flow provided in an embodiment of the present application;
fig. 8 is a schematic diagram of an optimal individual sharing process of the artificial fish swarm algorithm and the firefly algorithm provided in the embodiment of the present application.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
For convenience of understanding, referring to fig. 1, an embodiment of an operation and maintenance strategy generation method for a power transformer provided in the present application includes:
In the embodiment, state quantity data of 6 parts, namely a power transformer body, a sleeve, a tap switch, a cooling system, a non-electric quantity protection system and an online monitoring device, are mainly acquired; the transformer basic information mainly refers to information such as equipment price, load state or importance degree of the power transformer, importance degree and fault condition of equipment relative to a power grid and the like, and is mainly used for extracting deeper representative information; other information may also be selected according to actual situations, and is not limited herein.
In addition, the state quantity data of the power transformer body mainly comprises operation conditions, inspection and inspection, high-voltage tests, chromatographic analysis and simplified tests. The operation condition can be divided into the conditions of short circuit, transformer overload, neutral point direct current and the like; the inspection and inspection mainly comprises the steps of inspecting oil leakage, a jointing clamp, a body oil level, a tapping switch oil level, noise and vibration, foundation sinking, operating oil temperature, gas relay action, an oil storage cabinet, a breather, a secondary terminal, corrosion and iron core insulation; the high-voltage test mainly comprises infrared detection, direct-current resistance of the winding and the sleeve, dielectric loss of the winding, deformation test of the winding and insulation resistance test; chromatographic analysis including total hydrocarbons, acetylene and hydrogen; the simplified test comprises the furfural content in oil, the dielectric loss factor of the oil, the breakdown voltage of the oil, the moisture content, the gas content in the oil, the granularity test in the oil and the corrosive sulfur.
The sleeve mainly comprises a porcelain sleeve, an appearance, a test and a sleeve current transformer; the porcelain bushing is divided into creepage, external insulation configuration, porcelain insulation damage, composite external insulation cracking and hydrophobic performance; the appearance comprises oil level indication, oil leakage, bushing wiring and end screen outgoing lines; the test comprises infrared detection, dielectric loss and capacitance, partial discharge live test, insulating oil and insulating resistance; the bushing current transformer comprises oil leakage, direct current resistance and insulation resistance.
The tap changer can be divided into appearance, maintenance and test, and the appearance mainly comprises a tap position, an oil tank and oil leakage; the maintenance mainly refers to the switching times or the maintenance interval, a transmission mechanism, a filter element and a light gas relay inspection and control loop; the test mainly refers to dynamic characteristics, direct current resistance, oil breakdown voltage, oil chromatography detection, water content and oil withstand voltage.
The cooling system comprises a submersible pump, a cooling fan, a radiator, a cooler and the like; the oil-submersible pump comprises oil leakage, refusal to move and motor operation; the cooling fan comprises a movement rejection part, fan blades and abnormal sound; the radiator comprises oil leakage and a radiator radiating effect; the cooler comprises an oil leakage oil and a cooler electric valve; others include oil flow sensors, water flow sensors, and air pumps.
The non-electric quantity protection system comprises a cold control box, a gas relay, a thermometer, a pressure release valve, a pressure regulating switch oil flow (pressure) relay, a quick-acting oil pressure relay and a discharge gap; the cold control box comprises a contactor, a power supply, a cooling system overhaul, a cooler control system and a starting mode; the gas relay comprises oil leakage, window glasses, relay action, a rainproof cover and periodic verification; the thermometer comprises a temperature indication and a regular verification; the pressure relief valve comprises oil leakage, action indication, a secondary circuit and periodic verification; the pressure regulating switch oil flow relay comprises an oil leakage oil and a window mirror; the quick-acting oil pressure relay comprises oil leakage and relay action; the discharge gap includes inspection and inspection.
The on-line monitoring device comprises a transformer oil chromatogram on-line monitoring and measuring device, a sleeve on-line monitoring device, a neutral point direct current monitoring device, a leakage detecting device and other monitoring devices; the state quantity information comprises two types of state quantity data, namely detection data and device data.
And 102, extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence.
Further, step 102 includes:
calculating the deduction value of each component according to the state quantity data, and extracting the deduction value of each component based on preset weight;
performing state evaluation on each part according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating the fault probability of the equipment by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on equipment price, load importance and equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
It should be noted that, corresponding deduction standards are configured for all the state quantity information, the basic deduction value of the sub-item corresponding to each component can be determined according to the deduction standards, and then, the weight deduction value can be obtained by carrying out deduction calculation according to the weight coefficient. That is, there is a corresponding deduction value for each component's sub-item; a plurality of deduction values exist in each part, and the total deduction value of the part can be counted; even the total deduction value of the equipment can be counted for subsequent data analysis and calculation.
The process of evaluating the state of each component according to the deduction value and the preset evaluation rule can be described as follows: when the single deduction of any state quantity and the total deduction of the components are within a normal state range of a preset evaluation rule at the same time, the state quantity is regarded as a normal state; when the deduction of any state quantity single item or the deduction of all the state quantities of the components reaches the attention state range specified by the preset evaluation rule, the components are regarded as the attention state; when any single deduction of the state quantity reaches the abnormal state or serious state range specified by the preset evaluation rule, the state quantity is regarded as the abnormal state or serious state; the specific state region rule expression is shown in fig. 3.
The calculation of the mean failure probability of a device is expressed as:
wherein, l is a state evaluation result, the value of 1-4,1 represents a normal state, 2 represents an attention state, 3 represents an abnormal state, and 4 represents a severe state; p is l I.e. the component failure probability in a certain evaluation state, i.e. the mean failure probability of the device. n is a radical of an alkyl radical l 、N l The number of faulty devices and the total number of devices in a certain evaluation state are respectively.
Specifically, firstly, a defect index is defined to describe parameters of the severity of the part defect, and the higher the defect index is, the more serious the part defect is; the index is obtained by converting the state grade and the deduction value obtained by evaluating the components in the state evaluation process. The calculation process of the equipment defect index d is as follows:
wherein S is i Is an index of the evaluation state of the part, M is the actual total score of the equipment, M i The highest deduction value of the lower grade of the part. S i Is evaluated according to the state of the componentAnd the change is 0 in the normal state, 1 in the attention state, 2 in the abnormal state, and 3 in the serious state. M i The value of (2) is obtained by statistics according to the state evaluation result of the equipment.
In addition, the equipment fault probability P and the equipment defect index d obey an index distribution rule, in the actual operation process, the defect index is divided into a plurality of sections to be carried out, and the number of the equipment with faults in each defect index section in the target area is counted; the mean value of each defect index interval is selected to calculate the average fault probability of the interval, so that the fault probability of the equipment is represented. Theoretically, dividing more intervals will obtain more fitting points and more accurate results, but the premise is that there are enough sample data. In order to correspond to 4 status levels (normal, attention, abnormality, severity) of the status evaluation results, the defect index sections are divided into four sections of 0-1, 1-2, 2-3, and 3-4. The defect index-equipment failure probability is shown in table 1.
TABLE 1 Defect index-Equipment failure probability List
| Status rating | Index of defect | Number of faulty devices | Total number of equipment | Probability of failure |
| Is normal | 0≤d≤1 | n1 | N1 | n1/N1 |
| Attention is paid to | 1<d≤2 | n2 | N2 | n2/N2 |
| Abnormality (S) | 2<d≤3 | n3 | N3 | n3/N3 |
| Severe severity of |
3<d≤4 | n4 | N4 | n4/N4 |
The device value, the load level and the device status can be obtained by directly counting according to the device price, the load importance and the device importance, for example, the device value is described in a level form, please refer to table 2. The load grades can also be described according to three grades divided by the load importance degree, and specifically include three-grade load, two-grade load and one-grade load. The equipment status is divided according to the importance of the substation where the equipment is located in a power grid, and can be divided into a hub substation, an important substation and a general substation, and whether the requirement of N-1 is met according to the grid structure of the substation is considered; the operation mode of the power grid governed by each power supply bureau is dynamically determined every year, and the equipment status can also be subjected to grade quantization, specifically referring to table 3.
TABLE 2 device value rank List
| Value of the apparatus | Device class value range |
| Less than 10 ten thousand |
1 |
| More than 10 ten thousand yuan and less than 20 ten thousand yuan | 2 |
| 20 ten thousand yuan or more and 30 ten thousand yuan or less | 3 |
| More than 30 ten thousand yuan and less than 50 ten thousand yuan | 4 |
| More than 50 ten thousand yuan and less than 80 ten thousand |
5 |
| Over 80 ten thousand yuan and below 100 ten thousand |
6 |
| Over 100 ten thousand yuan and below 1000 ten thousand yuan | 7 |
| Over 1000 ten thousand yuan and below 5000 ten thousand yuan | 8 |
| More than 5000 ten thousand yuan and less than 1 hundred yuan | 9 |
| 1 hundred million yuan or more | 10 |
TABLE 3 device status rank List
Calculating the equipment loss degree based on equipment cost, personal safety factors and electric power safety factors:
wherein j takes 1-3, the equipment cost is represented if the value is 1, the personal safety is represented if the value is 2, and the power safety is represented if the value is 3; w Lj For the weight of the loss degree caused by the loss factor, the reference value is as follows: 1-equipment cost, taking 0.3; 2-personal safety, taking 0.3; 3-electric power safety, taking 0.4; l is j Degree of loss being a factor; degree of loss L of a certain factor j Calculated as follows:
k = 1-n, the loss grade of the loss factor and the equipment cost factor are represented, and n is 9; taking n as a personal safety factor to be 9; the power safety factor, n is 9; IOF jk The loss value of the loss factor under a certain grade; POF jk Is the probability of loss factor occurrence under a certain grade; the probability of the loss factor occurring is calculated as follows:
wherein j is a loss factor; k is the loss grade of the loss factor; n is the total failure times; n is jk The number of failures that is a loss factor at a certain level. The loss factors, i.e., the degree of equipment loss and the value range, are described in table 4. Loss factorThe number of failures can be obtained from statistics, i.e. the number of failures of each loss factor in different accident event classes.
Table 4 grading list corresponding to equipment loss degree
The operation and maintenance strategy is a set of operation and maintenance strategies of each component, and mainly comprises a daily operation and maintenance strategy, a special operation and maintenance strategy and a power failure operation and maintenance strategy. For example, the operation and maintenance times are specified in the operation and maintenance strategy to be 1 month and 1 time, 1 quarter and 1 time, half year and 1 year; define 1 month 1 as 1,1 quarterly 1 as 2 and half year 1 as 3,1 year 1 as 4.
The comprehensive data information can be constructed by integrating all the information, or the comprehensive data information is called as an operation and maintenance data set, and the operation and maintenance data set can be divided into 8; the training set is used for training the model, the verification set is used for carrying out parameter tuning on the model, and the test set is used for evaluating the performance effect of the model.
The embodiment converts the one-dimensional data information, namely the comprehensive data information into the two-dimensional image form, so that the data expression capacity can be enhanced, and the data analysis of the model is facilitated. The operation and maintenance strategy can be set to meet periodicity, one period is one year or half a year, and time is taken as a transverse variable factor to form a time sequence; taking the comprehensive data information as a longitudinal variable factor, so that a periodic two-dimensional image can be formed, and a two-dimensional image training data set is obtained; the verification set and the test set can also be operated in the same way, and data information with the same format is formed, so that unified processing is facilitated.
Further, converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set, wherein the method comprises the following steps:
and carrying out exception removal, normalization and one-hot coding processing on the comprehensive data information to realize data information standardization.
It should be noted that, because the dimensions of the integrated data information in this embodiment are not uniform and there are many information layers, some preprocessing operations, such as abnormal data removal, dimension normalization processing, and one-hot encoding, may also be performed before the data information is processed, so that the data information can be converted into standardized data, which is convenient for subsequent analysis.
For the operation objects of exception culling, such as missing values, the missing values are data exceptions generally caused by system temporary faults or storage system write errors, mainly include constant values or Null value data, and are easy to identify. For data loss in a long time period, directly discarding the sequence segment; for short-time data loss, the data loss is filled and corrected based on data characteristic analysis.
The normalization operation in this embodiment is to map different types of integrated data information between [0,1], which may be specifically expressed as:
wherein x is the original data, x max 、x min Maximum and minimum values, x, representing each type of information * Is the data obtained after normalization.
In the embodiment, a one-hot coding mechanism is adopted to code the normalized data, and all data information is subjected to binary vectorization expression, so that different data volumes are quantitatively distinguished, the data difference can be better expressed, and the data analysis of a model is facilitated; referring to fig. 4, after the integrated data information is used as an initial set and is subjected to standardization, dimension conversion may be performed, so as to obtain a two-dimensional image training data set.
And 103, training a preset convolutional neural network model by adopting a two-dimensional image training data set to obtain an optimized convolutional neural network model, and optimally training the parameters of the optimized model of the preset convolutional neural network model according to an artificial fish school-firefly algorithm to obtain the optimized convolutional neural network model.
The preset convolutional neural network model in the embodiment can be understood as a pre-built model with determined model parameters, and the model is trained through a two-dimensional image training data set, so that the performance of the model can be optimized, the analysis capability and the expression capability of the model on the operation and maintenance strategy related data are improved, and a prediction result which is more in line with the actual situation is obtained. The model parameters in the preset convolutional neural network model in the embodiment are obtained by optimizing an improved fusion algorithm, namely an artificial fish school-firefly algorithm, namely an optimization algorithm after the artificial fish school algorithm and the firefly algorithm are fused; because the population evolutions of the two optimization algorithms are synchronously performed, the optimal individual sharing of the population can be realized, the joint optimization iteration is realized, and the global search capability of the population is improved; therefore, the optimized model parameters obtained based on the optimization of the fusion algorithm can improve the learning capability of the model to a certain extent and ensure the training effect of the model.
Further, step 103, before, further includes:
constructing an initial convolutional neural network model based on an attention mechanism;
performing individual sharing-based joint optimization training on initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and adjusting the initial convolutional neural network model by optimizing the model parameters to obtain a preset convolutional neural network model.
The loss function of the initial convolutional neural network model in this embodiment is a central loss function, and may be expressed as:
wherein x is i As input feature vectors, c y Is the center of the class y to which x corresponds, c y Each iterative update is implemented by the mini-batch. The mini-batch is a parameter updating method for small batch gradient descent. In order to avoid large disturbance caused by a few sample marking errors, the embodiment adopts a scalar alpha c To controlc y Is a of c ∈[0,1]. In addition, in order to alleviate the overfitting of the model, dropout is adopted in the embodiment to reduce the dependency relationship between neurons, and the convergence speed of the model is increased.
In this embodiment, an Attention (Attention) mechanism is introduced, and two Attention-force mechanisms (Attention-left, attention-right) are used to perform Attention calculation on the output vector of the convolutional neural network layer and its internal features, so as to enhance the weights of the important time step vectors and the important features in each time step vector, thereby improving the precision of the operation and maintenance strategy selection, and please refer to fig. 5 for an example of the principle of the Attention mechanism.
Referring to fig. 6 and 7, the network structure of the initial convolutional neural network model in this embodiment mainly includes 9 convolutional layers, 3 pooling layers, and 3 full-link layers. The number of filters in the first convolutional layer is 256, and finally the output size is 2 through the full link layer. Except for the Sigmoid activation function used by the last fully-connected layer, the Relu function is used by the other convolutional layers and fully-connected layers, and the Flatten is used for one-dimensional operation between the last maximum pooling layer and the fully-connected layer. In addition, the network parameters of the initial convolutional neural network model of the present embodiment are shown in table 5.
TABLE 5 network parameter List for initial convolutional neural network model
| Serial number | Network layer | Filter size | Step size | Number of filters | Activating a |
| 1 | |
3×3 | 1×1 | 256 | Relu |
| 2 | Convolutional layer 2 | 3×3 | 1×1 | 128 | |
| 3 | |
3×3 | 1×1 | 128 | Relu |
| 4 | |
2×2 | 2×2 | / | / |
| 5 | Convolutional layer 4 | 3×3 | 1×1 | 128 | |
| 6 | |
3×3 | 1×1 | 128 | Relu |
| 7 | |
3×3 | 1×1 | 128 | Relu |
| 8 | Maximum pooling layer 2 | 2×1 | 2×1 | / | / |
| 9 | Convolutional layer 7 | 3×3 | 1×1 | 64 | |
| 10 | Convolutional layer 8 | 3×3 | 1×1 | 64 | Relu |
| 11 | Convolution layer 9 | 3×3 | 2×2 | 16 | |
| 12 | |
2×2 | 2×2 | / | / |
In the process of performing the individual sharing-based joint optimization training on the initial model parameters in the initial convolutional neural network model by adopting the artificial fish swarm-firefly algorithm, the most core is the fusion training of the artificial fish swarm algorithm and the firefly algorithm, and the two algorithms have certain advantages in solving specific problems, but still show the problems of insufficient adaptability, too slow convergence, local optimum and the like under certain conditions. Therefore, the embodiment sets that the optimal individuals of one population in one algorithm are shared to one population in another algorithm at regular intervals, and improves the global searching capability of the two populations.
Specifically, the initialized population parameters of the two algorithms can be set to be consistent, and the initial model parameters are the initialized population. In addition, the fitness values of the two types of populations can be calculated based on the central loss function in the convolutional neural network model in the embodiment; and respectively recording the global optimal positions searched by the two populations, and sequencing the individual search results from good to bad according to the fitness value to form a sequencing set. Observing whether algebraic vision needs to be shared, and if so, sharing the optimal individuals; the first m results and the last m results of the respective fitness ordered sets of the artificial fish school algorithm and the firefly algorithm are taken out and exchanged with each other, please refer to fig. 8. In order to prevent the situation that the value is not updated and the original path is returned in the subsequent iteration, the initial value of m is set to be not more than half of the population number. It may also fall into a "premature" condition as the number of iterations increases, so m is set to increase as the number of iterations increases:
wherein N is the population number, t and t max Current iteration number and maximum iteration number, r 0 For the proportional control coefficient, it can be set in the range of 0.51,1]In the interior of said container body,
is a rounding down function. During the exchange, the firefly algorithm population and the artificial fish swarm algorithm population update the position and the fixness value. And recording the global optimal value including the artificial fish swarm algorithm and the firefly algorithm. And judging whether a condition of stopping circulation is reached, if so, outputting a global optimal value and a global optimal position.
And 104, inputting the currently acquired real-time comprehensive data information into an optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
The currently acquired real-time comprehensive data information is data of the power transformer which needs to be researched and analyzed currently, and operation and maintenance strategies corresponding to components of the power transformer are not clear; the comprehensive data information acquired in real time is used as input, the predicted optimized operation and maintenance strategy is used as output, and the method has high-efficiency information processing capacity. In addition, it can be understood that the currently acquired real-time comprehensive data information also needs to be subjected to the data processing process, and is integrated into a data format accepted by the model, and the specific process is not repeated.
In addition, the daily operation and maintenance strategy of the power transformer is the minimum value of the corresponding numerical values of the daily operation and maintenance strategies of all the components of the power transformer; the special maintenance patrol strategy of the power transformer is a set formed by special maintenance patrol strategies of all parts of the power transformer; the power failure maintenance strategy of the power transformer is a set formed by power failure maintenance strategies of all parts of the power transformer.
According to the operation and maintenance strategy generation method of the power transformer, operation and maintenance data analysis is carried out by acquiring state quantity data of different parts of the power transformer and transformer basic information, and then an optimized neural network model is trained according to comprehensive data information formed by various parts and equipment data, so that the model has stronger expression capacity in operation and maintenance strategy generation, and the accuracy of the generated optimized operation and maintenance strategy can be ensured; the model parameters are optimized and trained by adopting a specific algorithm, so that the learning performance of the model can be improved, and the reliability of the model is further ensured. Therefore, the technical problems that in the prior art, operation and maintenance strategies cannot be designed according to the component health state of the power transformer, the actually obtained operation and maintenance strategies lack pertinence, and the effect is poor can be solved.
For easy understanding, please refer to fig. 2, the present application provides an embodiment of an operation and maintenance strategy generation apparatus for a power transformer, including:
an information acquisition unit 201, configured to acquire state quantity data and transformer basic information of different components of a power transformer, respectively;
the information processing unit 202 is used for extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence;
the model training unit 203 is used for training a preset convolutional neural network model by adopting a two-dimensional image training data set to obtain an optimized convolutional neural network model, and the optimized model parameters of the preset convolutional neural network model are obtained by optimized training according to an artificial fish swarm-firefly algorithm;
and the strategy generating unit 204 is configured to input the currently acquired real-time comprehensive data information into the optimized convolutional neural network model for operation and maintenance strategy analysis, so as to obtain an optimized operation and maintenance strategy.
Further, the information processing unit 202 is specifically configured to:
calculating the deduction value of each component according to the state quantity data, and extracting the deduction value of each component based on preset weight;
performing state evaluation on each part according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating the equipment fault probability by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on equipment price, load importance and equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
Further, still include:
and the normalizing unit 205 is configured to perform exception removal, normalization and one-hot encoding on the comprehensive data information, so as to achieve data information normalization.
Further, still include:
a model construction unit 206, configured to construct an initial convolutional neural network model based on an attention mechanism;
the parameter optimization unit 207 is used for performing joint optimization training based on individual sharing on the initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and the model optimization unit 208 is configured to adjust the initial convolutional neural network model by optimizing the model parameters to obtain a preset convolutional neural network model.
The application also provides operation and maintenance strategy generation equipment of the power transformer, and the equipment comprises a processor and a memory;
the memory is used for storing the program codes and transmitting the program codes to the processor;
the processor is used for executing the operation and maintenance strategy generation method of the power transformer in the above method embodiment according to the instructions in the program code.
The present application further provides a computer-readable storage medium for storing program codes for executing the operation and maintenance strategy generation method of the power transformer in the above method embodiment.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for executing all or part of the steps of the method described in the embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device). And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. An operation and maintenance strategy generation method of a power transformer is characterized by comprising the following steps:
respectively acquiring state quantity data and transformer basic information of different parts of a power transformer;
extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information, and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence;
training a preset convolutional neural network model by using the two-dimensional image training data set to obtain an optimized convolutional neural network model, wherein the optimized model parameters of the preset convolutional neural network model are obtained by optimized training according to an artificial fish swarm-firefly algorithm;
and inputting the currently acquired real-time comprehensive data information into the optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
2. The method for generating an operation and maintenance strategy of a power transformer according to claim 1, wherein the extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information and constructing a two-dimensional image training data set according to the comprehensive data information and a time sequence comprises:
calculating a withholding value of each component according to the state quantity data, and extracting the withholding value of each component based on preset weight;
performing state evaluation on each part according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating the fault probability of the equipment by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on equipment price, load importance and equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
3. The method for generating the operation and maintenance strategy of the power transformer according to claim 2, wherein the step of converting the one-dimensional data information into the two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set further comprises the steps of:
and carrying out exception removal, normalization and one-hot coding processing on the comprehensive data information to realize data information standardization.
4. The method for generating an operation and maintenance strategy of a power transformer according to claim 1, wherein the training of the preset convolutional neural network model by using the two-dimensional image training data set is performed to obtain an optimized convolutional neural network model, and the parameters of the optimized model of the preset convolutional neural network model are obtained by performing optimization training according to an artificial fish swarm-firefly algorithm, and the method further comprises the following steps:
constructing an initial convolutional neural network model based on an attention mechanism;
performing individual sharing-based joint optimization training on initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and adjusting the initial convolutional neural network model through the optimization model parameters to obtain a preset convolutional neural network model.
5. An operation and maintenance strategy generation device of a power transformer is characterized by comprising the following steps:
the information acquisition unit is used for respectively acquiring state quantity data and transformer basic information of different parts of the power transformer;
the information processing unit is used for extracting comprehensive data information of the power transformer according to the state quantity data and the transformer basic information and constructing a two-dimensional image training data set according to the comprehensive data information and the time sequence;
the model training unit is used for training a preset convolutional neural network model by adopting the two-dimensional image training data set to obtain an optimized convolutional neural network model, and the optimized model parameters of the preset convolutional neural network model are obtained by optimized training according to an artificial fish swarm-firefly algorithm;
and the strategy generating unit is used for inputting the currently acquired real-time comprehensive data information into the optimized convolutional neural network model for operation and maintenance strategy analysis to obtain an optimized operation and maintenance strategy.
6. The operation and maintenance strategy generation device of the power transformer according to claim 5, wherein the information processing unit is specifically configured to:
calculating a withholding value of each component according to the state quantity data, and extracting the withholding value of each component based on preset weight;
performing state evaluation on each component according to the deduction value and a preset evaluation rule to obtain a state evaluation result, and calculating equipment fault probability by combining the state evaluation result and the deduction value;
respectively acquiring equipment value, load grade and equipment status based on the equipment price, the load importance and the equipment importance;
calculating the equipment loss degree based on the equipment cost, the personal safety factor and the electric power safety factor;
constructing comprehensive data information according to the equipment fault probability, the equipment value, the load grade, the equipment status, the equipment loss degree and a preset operation and maintenance strategy;
and converting the one-dimensional data information into a two-dimensional image according to the comprehensive data information and the time sequence to obtain a two-dimensional image training data set.
7. The operation and maintenance strategy generation device of the power transformer according to claim 6, further comprising:
and the standardization unit is used for carrying out exception removal, normalization and one-hot coding processing on the comprehensive data information to realize data information standardization.
8. The operation and maintenance strategy generation device of the power transformer according to claim 5, further comprising:
the model building unit is used for building an initial convolutional neural network model based on an attention mechanism;
the parameter optimization unit is used for performing joint optimization training based on individual sharing on the initial model parameters in the initial convolutional neural network model by adopting an artificial fish swarm-firefly algorithm to obtain optimized model parameters;
and the model optimization unit is used for adjusting the initial convolutional neural network model through the optimization model parameters to obtain a preset convolutional neural network model.
9. An operation and maintenance strategy generation device of a power transformer is characterized by comprising a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is used for executing the operation and maintenance strategy generation method of the power transformer according to any one of claims 1-4 according to instructions in the program code.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing program code for executing the operation and maintenance strategy generation method of the power transformer of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211591750.1A CN115796843A (en) | 2022-12-12 | 2022-12-12 | Operation and maintenance strategy generation method of power transformer and related device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211591750.1A CN115796843A (en) | 2022-12-12 | 2022-12-12 | Operation and maintenance strategy generation method of power transformer and related device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115796843A true CN115796843A (en) | 2023-03-14 |
Family
ID=85419354
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211591750.1A Pending CN115796843A (en) | 2022-12-12 | 2022-12-12 | Operation and maintenance strategy generation method of power transformer and related device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115796843A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117951588A (en) * | 2024-03-27 | 2024-04-30 | 南方电网科学研究院有限责任公司 | Modeling method, device, terminal and medium for OIP sleeve health state evaluation model |
-
2022
- 2022-12-12 CN CN202211591750.1A patent/CN115796843A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117951588A (en) * | 2024-03-27 | 2024-04-30 | 南方电网科学研究院有限责任公司 | Modeling method, device, terminal and medium for OIP sleeve health state evaluation model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106980905B (en) | Power distribution network power supply reliability prediction method and system | |
| CN109711663A (en) | Substation's oil-immersed transformer status assessment and modification method and system based on big data analysis | |
| CN113779496A (en) | Power equipment state evaluation method and system based on equipment panoramic data | |
| Yang et al. | State evaluation of power transformer based on digital twin | |
| CN106651189B (en) | Transformer state evaluation method based on multilayer composite rule | |
| CN109583520B (en) | State evaluation method of cloud model and genetic algorithm optimization support vector machine | |
| CN115564203A (en) | Equipment real-time performance evaluation system and method based on multi-dimensional data cooperation | |
| CN112163371B (en) | Transformer bushing state evaluation method | |
| CN110879377B (en) | Metering device fault tracing method based on deep belief network | |
| Jin-qiang | Fault prediction of a transformer bushing based on entropy weight TOPSIS and gray theory | |
| CN113313371B (en) | Power distribution network risk assessment method, device, computer equipment and storage medium | |
| CN113343581B (en) | Transformer fault diagnosis method based on graph Markov neural network | |
| CN113435759B (en) | Primary equipment risk intelligent assessment method based on deep learning | |
| CN113987294A (en) | CVT (continuously variable transmission) online fault diagnosis method based on genetic optimization GRU (generalized regression Unit) neural network | |
| CN110688809A (en) | Box transformer substation fault diagnosis method based on VPRS-RBF neural network | |
| CN115796843A (en) | Operation and maintenance strategy generation method of power transformer and related device | |
| CN110705859A (en) | PCA-self-organizing neural network-based method for evaluating running state of medium and low voltage distribution network | |
| CN114462508A (en) | Power transformer health state assessment method based on multi-mode neural network | |
| CN113342784A (en) | Database design method for risk assessment of main transformer equipment of power grid | |
| CN115689114A (en) | Submarine cable running state prediction method based on combined neural network | |
| CN116562114A (en) | Power transformer fault diagnosis method based on graph convolution neural network | |
| CN115271238A (en) | Method and device for predicting service life of transformer | |
| CN115511387A (en) | Oil immersed transformer state evaluation method and system based on multi-source information fusion | |
| CN113985156A (en) | Intelligent fault identification method based on transformer voiceprint big data | |
| CN117332330A (en) | Transformer fault evaluation method and system based on multi-source data graph fusion model |
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 |