Intelligent control power distribution cabinet fault diagnosis method and system based on big data

Abstract

The invention discloses an intelligent control power distribution cabinet fault diagnosis method and system based on big data, and relates to the technical field of power distribution cabinet fault diagnosis, and the method comprises the steps: S1, data collection, S2, data preprocessing, S3, data fusion, S4, feature extraction, S5, fault diagnosis, and S6, fault alarm. According to the method, data of various sources and different structures are integrated, through data fusion, the data of different types and different sources are effectively fused, and hidden association relationships among the data are mined, so that the operation state of the power distribution cabinet is reflected more comprehensively and accurately, the precision and reliability of fault diagnosis are improved, and the fault diagnosis efficiency is improved. Particularly, the insulation performance of electrical elements in the power distribution cabinet may be affected in a high-temperature and humid environment, and potential faults caused by environmental factors can be predicted in advance by combining environmental data and electrical data instead of being detected through electrical parameter abnormalities after the faults occur.

Description

Technical Field

[0001] The present invention belongs to the field of power distribution cabinet fault diagnosis, and in particular relates to a power distribution cabinet fault diagnosis method and system based on big data intelligent control. Background Art

[0002] The power distribution cabinet is an important device used to distribute electric energy, control circuits and protect equipment in the power system. Its safe and stable operation is crucial to the reliability of the power system. As the scale of power systems continues to expand and the complexity increases, the fault diagnosis of power distribution cabinets faces increasing challenges.

[0003] Traditional fault diagnosis methods for distribution cabinets are mainly based on manual inspections and experience-based judgments, which have problems such as low diagnostic efficiency, poor accuracy, and lack of timeliness. Manual inspections make it difficult to monitor the operating status of distribution cabinets in real time, and problems can often only be discovered after a fault occurs, resulting in prolonged power outages and affecting production and life. Moreover, manual judgment of the cause of the fault is easily affected by subjective factors, which may lead to misdiagnosis or missed diagnosis.

[0004] With the rapid development of big data technology, big data technology can collect and store a large amount of operating data of distribution cabinets in real time. Through data analysis and mining, early warning and accurate diagnosis of distribution cabinet faults can be achieved. However, most distribution cabinet fault diagnosis systems currently rely mainly on the electrical parameter data inside the distribution cabinet for fault judgment, and are unable to diagnose potential faults caused by environmental factors, resulting in misjudgment or missed judgment. In response to the above problems, the following solutions are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a distribution cabinet fault diagnosis method and system based on big data intelligent control, which integrates data from multiple sources and with different structures to more comprehensively and accurately reflect the operating status of the distribution cabinet, thereby solving the existing problem of being unable to diagnose potential faults caused by environmental factors.

[0006] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] The present invention is a distribution cabinet fault diagnosis method based on big data intelligent control, comprising:

[0008] Step S1, data collection: collect various data of the power distribution cabinet at predetermined time intervals, and at the same time, regularly obtain relevant heterogeneous data from the historical maintenance database and peripheral devices;

[0009] Step S2, data preprocessing: cleaning and standardizing the collected data;

[0010] Step S3, data fusion: fusing the preprocessed data to obtain comprehensive operation data;

[0011] Step S4, feature extraction: extract features from the comprehensive data and combine the features into a feature vector as the input of the fault diagnosis model;

[0012] Step S5, fault diagnosis: construct a fault diagnosis model, and use historical fault data and normal operation data to pre-train and fine-tune the fault diagnosis model;

[0013] Step S6, fault alarm: If the fault diagnosis model determines that there is a fault in the power distribution cabinet, a corresponding alarm mechanism is triggered according to the severity of the fault.

[0014] Preferably, the step S2, data preprocessing, specifically comprises the following steps:

[0015] Step S21: Smoothing the electrical parameters with large fluctuations. n , the moving average value y t The calculation formula is:

[0016]

[0017] In the formula, m is the window size of the moving average, x i is the data point in the original time series, i is an index variable used to traverse all data points from t-m+1 to t;

[0018] At the same time, obviously erroneous data points are identified and corrected based on the normal operating range of the sensor and the physical characteristics of the data;

[0019] Step S22: Standardize different types of data to normalize electrical parameters and environmental parameters to the same value range. i , its standardized value z i The normalization formula is:

[0020]

[0021] In the formula, μ is the mean of the data set, and σ is the standard deviation of the data set.

[0022] Preferably, the step S3, data fusion, specifically comprises the following steps:

[0023] Step S31: For each data source, determine the initial credibility distribution function m according to its data characteristics and historical reliability i , where i = 1, 2, Λ, n, calculate the credibility weight of each data source;

[0024] Step S32: Perform data fusion to obtain a fused basic credibility distribution function, and obtain comprehensive data reflecting the operating status of the distribution cabinet.

[0025] Preferably, the credibility weight calculation formula in step S31 is:

[0026]

[0027] In the formula, k i is the initial weight coefficient of data source i, d i is the data uncertainty measure of data source i, and λ is the adjustment parameter.

[0028] Preferably, the basic credibility distribution function in step S32 is:

[0029]

[0030] In the formula, m f (A) is the value of the basic credibility distribution function for event A after fusion, A and A i is a subset of the running state, φ represents an empty set, n is the number of data sources, ω i is the credibility weight of the i-th data source;

[0031] Preferably, the step S5, fault diagnosis specifically comprises the following steps:

[0032] Step S51: pre-training a deep learning model on a large-scale general distribution cabinet fault data set to learn general fault modes and feature representations;

[0033] Step S52: define the difference feature vector ΔX between the specific distribution cabinet data and the pre-training data:

[0034] ΔX=X spec -X pre ;

[0035] Where, X pre Output features of the feature extraction layer of the pre-trained model, X spec Output features for specific distribution cabinet data through the same feature extraction layer;

[0036] Step S53: ΔX is processed by a difference feature enhancement network DFEN, and the output of DFEN is ΔX enhanced , the calculation formula is as follows:

[0037]

[0038] Where W i and b iare the weight matrix and bias vector of the i-th layer respectively, and σ is the activation function;

[0039] Step S54: Fuse the enhanced difference features with the features of the pre-trained model to obtain the input features X of the fine-tuning model fine-tune :

[0040] Xfine-tune=Xpre+α×ΔXenhanced;

[0041] In the formula, α is the fusion coefficient.

[0042] Preferably, a distribution cabinet fault diagnosis system based on intelligent control of big data, the diagnosis system includes a data acquisition module, a data preprocessing module, a data fusion module, a feature extraction module, a fault diagnosis module and a fault alarm module, the data acquisition module, the data preprocessing module, the data fusion module, the feature extraction module, the fault diagnosis module and the fault alarm module are connected in sequence, the output end of the data acquisition module is unidirectionally connected to the input end of the fault diagnosis module, and the historical fault data acquired by the data acquisition module is also provided to the fault diagnosis module for model training and learning.

[0043] The present invention has the following beneficial effects:

[0044] 1. The present invention integrates data from multiple sources and with different structures. Through data fusion, these data of different types and sources are effectively integrated to mine the hidden correlations between the data, thereby more comprehensively and accurately reflecting the operating status of the distribution cabinet and improving the accuracy and reliability of fault diagnosis. For example, a high temperature and humid environment may affect the insulation performance of electrical components in the distribution cabinet. Combining environmental data with electrical data can predict potential faults caused by environmental factors in advance, rather than just detecting them through abnormal electrical parameters after the fault occurs.

[0045] 2. The present invention utilizes transfer learning technology to first perform pre-training on a large number of distribution cabinet data sets with common fault characteristics to learn common fault diagnosis knowledge and model structure, and then fine-tune the pre-trained model for a specific distribution cabinet using its small amount of labeled data to quickly adapt to the fault diagnosis needs of the specific distribution cabinet. This can reduce dependence on a large amount of fault data from a specific distribution cabinet, and in particular, for some newly installed distribution cabinets or those with less data accumulation, can achieve efficient and accurate fault diagnosis in a short time, thereby reducing the deployment cost and time cost of the fault diagnosis system and improving the versatility and adaptability of the system.

[0046] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0048] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0049] Figure 2 It is a system framework diagram of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] See also Figure 1 As shown, the present invention is a distribution cabinet fault diagnosis method based on big data intelligent control, comprising:

[0052] Step S1, data collection: collect various data of the power distribution cabinet at predetermined time intervals, and at the same time, regularly obtain relevant heterogeneous data from the historical maintenance database and peripheral devices;

[0053] Step S2, data preprocessing: cleaning and standardizing the collected data;

[0054] Step S3, data fusion: fusing the preprocessed data to obtain comprehensive operation data;

[0055] Step S4, feature extraction: extract features from the comprehensive data and combine the features into a feature vector as the input of the fault diagnosis model;

[0056] Step S5, fault diagnosis: construct a fault diagnosis model, and use historical fault data and normal operation data to pre-train and fine-tune the fault diagnosis model;

[0057] Step S6, fault alarm: If the fault diagnosis model determines that there is a fault in the power distribution cabinet, a corresponding alarm mechanism is triggered according to the severity of the fault.

[0058] Step S2, data preprocessing specifically includes the following steps:

[0059] Step S21: Smoothing the electrical parameters with large fluctuations.n , the moving average value y t The calculation formula is:

[0060]

[0061] In the formula, m is the window size of the moving average, x i is the data point in the original time series, i is an index variable used to traverse all data points from t-m+1 to t;

[0062] At the same time, obviously erroneous data points are identified and corrected based on the normal operating range of the sensor and the physical characteristics of the data;

[0063] Step S22: Standardize different types of data to normalize electrical parameters and environmental parameters to the same value range. i , its standardized value z i The normalization formula is:

[0064]

[0065] In the formula, μ is the mean of the data set, and σ is the standard deviation of the data set.

[0066] Step S3, data fusion specifically includes the following steps:

[0067] Step S31: For each data source, determine the initial credibility distribution function m according to its data characteristics and historical reliability i , where i = 1, 2, Λ, n, calculate the credibility weight of each data source;

[0068] Step S32: Perform data fusion to obtain a fused basic credibility distribution function, and obtain comprehensive data reflecting the operating status of the distribution cabinet.

[0069] The calculation formula of the credibility weight in step S31 is:

[0070]

[0071] In the formula, k i is the initial weight coefficient of data source i, d i is the data uncertainty measure of data source i, and λ is the adjustment parameter.

[0072] The basic credibility distribution function in step S32 is:

[0073]

[0074] In the formula, m f (A) is the value of the basic credibility distribution function for event A after fusion, A and Ai is a subset of the running state, φ represents an empty set, n is the number of data sources, ω i is the credibility weight of the i-th data source;

[0075] Step S5, fault diagnosis specifically includes the following steps:

[0076] Step S51: pre-training a deep learning model on a large-scale general distribution cabinet fault data set to learn general fault modes and feature representations;

[0077] Step S52: define the difference feature vector ΔX between the specific distribution cabinet data and the pre-training data:

[0078] ΔX=X spec -X pre ;

[0079] Where, X pre Output features of the feature extraction layer of the pre-trained model, X spec Output features for specific distribution cabinet data through the same feature extraction layer;

[0080] Step S53: ΔX is processed by a difference feature enhancement network DFEN, and the output of DFEN is ΔX enhanced , the calculation formula is as follows:

[0081]

[0082] Where W i and b i are the weight matrix and bias vector of the i-th layer respectively, and σ is the activation function;

[0083] Step S54: Fuse the enhanced difference features with the features of the pre-trained model to obtain the input features X of the fine-tuning model fine-tune :

[0084] Xfine-tune=Xpre+α×ΔXenhanced;

[0085] In the formula, α is the fusion coefficient.

[0086] See also Figure 2As shown, the present invention is a distribution cabinet fault diagnosis system based on intelligent control of big data. The diagnosis system includes a data acquisition module, a data preprocessing module, a data fusion module, a feature extraction module, a fault diagnosis module and a fault alarm module. The data acquisition module, the data preprocessing module, the data fusion module, the feature extraction module, the fault diagnosis module and the fault alarm module are connected in sequence. The output end of the data acquisition module is unidirectionally connected to the input end of the fault diagnosis module. The historical fault data acquired by the data acquisition module is also provided to the fault diagnosis module for model training and learning.

[0087] A specific application of this embodiment is:

[0088] Step S1, data acquisition: deploy various types of sensors at various key locations of the power distribution cabinet, including but not limited to current transformers, voltage transformers, temperature sensors, humidity sensors, vibration sensors, etc., to obtain electrical parameters, environmental parameters, and operating status parameters of the power distribution cabinet; these sensors transmit the collected data to the data preprocessing module in real time via wired or wireless communication;

[0089] At the same time, the system is also connected to the historical maintenance database of the power distribution cabinet and the operation data interface of the surrounding related equipment to obtain more heterogeneous data, such as the maintenance record of the power distribution cabinet, the time of component replacement, the working current and voltage of adjacent equipment, etc., to realize the collection of multi-source data;

[0090] Step S2: Data preprocessing:

[0091] Step S21: Clean the received data to remove noise and outliers; use data smoothing algorithm to smooth the electrical parameters with large fluctuations; for time series data x1, x2, Λ, x n , the moving average value y t The calculation formula is:

[0092]

[0093] In the formula, x i is the data point in the original time series, i is an index variable used to traverse all data points from t-m+1 to t, and m is the window size of the moving average, which eliminates the impact of instantaneous interference on the data;

[0094] At the same time, data verification technology is used to identify and correct obviously erroneous data points based on the normal operating range of the sensor and the physical characteristics of the data;

[0095] Step S22: Standardize different types of data to normalize electrical parameters and environmental parameters to the same value range for subsequent data fusion and analysis; for example, for each data point x in the data seti , its standardized value z i The normalization formula is:

[0096]

[0097] In the formula, μ is the mean of the data set, σ is the standard deviation of the data set, so that the mean of the data is 0 and the standard deviation is 1;

[0098] Step S3: Data fusion:

[0099] Step S31: Assume that after the data obtained from n different data sources (such as electrical parameter data source, environmental data source, historical maintenance data source, etc.) are preprocessed, for a certain operating status indicator of the distribution cabinet, each data source provides a basic credibility distribution function m i (i=1,2,Λ,n), indicating the degree of support of the data source for each possible operating state (such as normal, fault 1, fault 2, etc.);

[0100] First, calculate the credibility weight ω of each data source i :

[0101]

[0102] In the formula, k i is the initial weight coefficient of data source i, λ is the adjustment parameter used to control the influence of uncertainty on the weight, d i is the data uncertainty measure of data source i, for example, calculated by the variance of the data. Let the data sample of data source i be x i1 ,x i2 ,Λ,x im , and its variance calculation formula is: In the formula, is the mean of the data from data source i;

[0103] Step S32: Then, the improved evidence theory combination rule is used to perform data fusion to obtain the fused basic credibility distribution function m f :

[0104]

[0105] In the formula, m f (A) is the value of the basic credibility distribution function for event A after fusion, A and A i is a subset of the running state, φ represents an empty set, n is the number of data sources, ω i is the credibility weight of the i-th data source;

[0106] In this way, the fusion weight can be dynamically adjusted according to the uncertainty of the data and the importance of the data source, so as to more accurately judge the operating status of the distribution cabinet by integrating multi-source data;

[0107] Step S4, feature extraction: extract multiple features from the fused data, including time domain features (such as mean, variance, peak, kurtosis, etc.) and frequency domain features (such as spectrum peak, frequency component, etc.); for time domain features, taking current data I(t) as an example, the mean calculation formula is:

[0108]

[0109] In the formula, T is the time period of data collection;

[0110] At the same time, the autoencoder in deep learning is used to learn the features of the data and extract deep feature representations to enhance the expression ability of fault features and provide richer information for fault diagnosis;

[0111] Step S5: Fault diagnosis:

[0112] Step S51: constructing a fault diagnosis model based on transfer learning;

[0113] First, a deep learning model, such as a convolutional neural network (CNN) or a long short-term memory network (LSTM), is pre-trained on a large-scale common distribution cabinet fault dataset to learn common fault modes and feature representations.

[0114] Step S52: In the process of applying the pre-trained model to the fault diagnosis fine-tuning of a specific distribution cabinet, define the difference feature vector ΔX between the specific distribution cabinet data and the pre-trained data; let the feature extraction layer output feature of the pre-trained model be X pre , the output feature of the specific distribution cabinet data after the same feature extraction layer is X spec ,but:

[0115] ΔX=X spec -X pre ;

[0116] Step S53: Then, ΔX is processed by a difference feature enhancement network (DFEN). DFEN consists of multiple fully connected layers and activation functions. Let the output of DFEN be ΔX enhanced , and its calculation formula is as follows:

[0117]

[0118] Where W i and b i are the weight matrix and bias vector of the i-th layer respectively, σ is the activation function, and its formula is ReLU(x)=max(0,x);

[0119] Step S54: Finally, the enhanced difference features are fused with the features of the pre-trained model to obtain the input features X of the fine-tuning model. fine-tune :

[0120] Xfine-tune=Xpre+α×ΔXenhanced;

[0121] In the formula, α is the fusion coefficient, which is used to balance the contribution of pre-training features and difference features;

[0122] In this way, the knowledge of the pre-trained model can be better utilized and quickly adapted to the characteristic differences of specific distribution cabinets, improving the accuracy and efficiency of fault diagnosis;

[0123] Step S6, fault alarm:

[0124] According to the output results of the fault diagnosis model, determine whether there is a fault in the power distribution cabinet and the severity of the fault; for minor faults, display fault prompt information on the human-machine interface of the monitoring system, including the fault type, possible causes and recommended treatment measures;

[0125] For serious faults, in addition to being displayed on the interface, an alarm is sounded through an audible and visual alarm, and a text message notification is sent to the mobile phone of the relevant maintenance personnel, informing them of the fault details and the location of the distribution cabinet, so that maintenance measures can be taken in time to ensure the safe operation of the distribution cabinet.

[0126] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0127] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A distribution cabinet fault diagnosis method based on big data intelligent control, characterized in that: The diagnostic method comprises the following steps: Step S1, data collection: collect various data of the power distribution cabinet at predetermined time intervals, and at the same time, regularly obtain relevant heterogeneous data from the historical maintenance database and peripheral devices; Step S2, data preprocessing: cleaning and standardizing the collected data; Step S3, data fusion: fusing the preprocessed data to obtain comprehensive operation data; Step S4, feature extraction: extract features from the comprehensive data and combine the features into a feature vector as the input of the fault diagnosis model; Step S5, fault diagnosis: construct a fault diagnosis model, and use historical fault data and normal operation data to pre-train and fine-tune the fault diagnosis model; Step S6, fault alarm: When the fault diagnosis model determines that there is a fault in the power distribution cabinet, a corresponding alarm mechanism is triggered according to the severity of the fault.

2. According to the method of power distribution cabinet fault diagnosis based on big data intelligent control according to claim 1, it is characterized in that: The step S2, data preprocessing, specifically comprises the following steps: Step S21: Smoothing the fluctuating electrical parameters. n , the moving average value y t The calculation formula is: In the formula, m is the window size of the moving average, x i is the data point in the original time series, i is an index variable used to traverse all data points from t-m+1 to t; At the same time, obviously erroneous data points are identified and corrected based on the normal operating range of the sensor and the physical characteristics of the data; Step S22: Standardize different types of data to normalize electrical parameters and environmental parameters to the same value range. i , its standardized value z i The normalization formula is: In the formula, μ is the mean of the data set, and σ is the standard deviation of the data set.

3. The method for fault diagnosis of a power distribution cabinet based on intelligent control of big data according to claim 1 is characterized in that: The step S3, data fusion, specifically comprises the following steps: Step S31: For each data source, determine the initial credibility distribution function m according to its data characteristics and historical reliability i , where i = 1, 2, Λ, n, calculate the credibility weight of each data source; Step S32: Perform data fusion to obtain a fused basic credibility distribution function, and obtain comprehensive data reflecting the operating status of the distribution cabinet.

4. A method for diagnosing faults in a power distribution cabinet based on intelligent control of big data according to claim 3, characterized in that: The calculation formula of the credibility weight in step S31 is: In the formula, k i is the initial weight coefficient of data source i, d i is the data uncertainty measure of data source i, and λ is the adjustment parameter.

5. The method for fault diagnosis of a power distribution cabinet based on intelligent control of big data according to claim 1 is characterized in that: The basic credibility distribution function in step S32 is: In the formula, m f (A) is the value of the basic credibility distribution function for event A after fusion, A and A i is a subset of the running state, φ represents an empty set, n is the number of data sources, ω i is the credibility weight of the i-th data source.

6. The method for fault diagnosis of a power distribution cabinet based on intelligent control of big data according to claim 1 is characterized in that: The step S5, fault diagnosis specifically comprises the following steps: Step S51: pre-training a deep learning model on a large-scale general distribution cabinet fault data set to learn general fault modes and feature representations; Step S52: define the difference feature vector ΔX between the specific distribution cabinet data and the pre-training data: ΔX=X spec -X pre ; In the formula, X pre Output features of the feature extraction layer of the pre-trained model, X spec Output features for specific distribution cabinet data through the same feature extraction layer; Step S53: ΔX is processed by a difference feature enhancement network DFEN, and the output of DFEN is ΔX enhanced , the calculation formula is as follows: Where W i and b i are the weight matrix and bias vector of the i-th layer respectively, and σ is the activation function; Step S54: Fuse the enhanced difference features with the features of the pre-trained model to obtain the input features X of the fine-tuning model fine-tune : Xfine-tune=Xpre+α×ΔXenhanced; In the formula, α is the fusion coefficient.

7. A distribution cabinet fault diagnosis system based on big data intelligent control, characterized in that: The diagnostic system includes a data acquisition module, a data preprocessing module, a data fusion module, a feature extraction module, a fault diagnosis module and a fault alarm module. The data acquisition module, the data preprocessing module, the data fusion module, the feature extraction module, the fault diagnosis module and the fault alarm module are connected in sequence. The output end of the data acquisition module is unidirectionally connected to the input end of the fault diagnosis module. The historical fault data acquired by the data acquisition module is also provided to the fault diagnosis module for model training and learning.

Images