Intelligent analysis method and system for monitoring equipment in power distribution area

Abstract

The invention discloses a power distribution area equipment monitoring intelligent analysis method, and relates to the technical field of power system equipment monitoring, and the method comprises the steps: collecting and preprocessing the real-time operation data of power distribution area equipment; performing feature extraction on the real-time operation data by using a deep learning model to obtain an operation state feature vector; performing similarity calculation on the running state feature vector and a normal running state feature vector in a historical database to obtain a preliminary abnormal degree quantitative index; performing equipment state evaluation and classification according to the preliminary abnormal degree quantitative index to obtain an equipment state evaluation result; and taking corresponding measures according to an equipment state evaluation result, automatically generating a fault diagnosis report and pushing the fault diagnosis report to a maintenance terminal. The multi-parameter monitoring and deep learning feature extraction technology adopted by the invention improves the accuracy and comprehensiveness of anomaly detection, and can identify some potential fault features which are difficult to find by a traditional method.

Description

Technical Field

[0001] The present invention relates to the technical field of power system equipment monitoring, and in particular to an intelligent analysis method and system for monitoring equipment in a distribution station area. Background Art

[0002] The distribution area is an important link between the substation and the user in the power system, and its operating status directly affects the quality and reliability of power supply. With the development of smart grids, higher requirements are placed on the status monitoring and fault diagnosis of distribution area equipment. Traditional distribution area equipment monitoring methods mainly rely on regular inspections and simple threshold alarms, which have the following problems: First, data collection is not comprehensive, often only focusing on basic parameters such as voltage and current, and it is difficult to fully reflect the status of the equipment; second, the sampling frequency is low, and it is difficult to capture instantaneous anomalies; third, the data analysis capability is limited, and complex fault modes cannot be effectively identified; finally, fault diagnosis mainly relies on manual experience, which is inefficient and easily affected by subjective factors. These problems make it difficult to detect abnormal equipment status in a timely manner, increasing the risk of equipment failure and maintenance costs. Summary of the invention

[0003] In view of the problems existing in the existing method for monitoring equipment in a distribution station area, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is how to improve the accuracy and comprehensiveness of equipment status monitoring in a distribution station area, while realizing intelligent fault diagnosis and optimal allocation of maintenance resources.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In the first aspect, an embodiment of the present invention provides an intelligent analysis method for monitoring equipment in a distribution substation area, which includes collecting and preprocessing real-time operating data of equipment in a distribution substation area; the real-time operating data includes voltage, current, power factor, and temperature parameters; using a deep learning model to extract features of the real-time operating data to obtain an operating state feature vector; calculating the similarity between the operating state feature vector and the normal operating state feature vector in a historical database to obtain a preliminary abnormality degree quantitative index; evaluating and classifying the equipment state according to the preliminary abnormality degree quantitative index to obtain an equipment state evaluation result; taking corresponding measures according to the equipment state evaluation result, automatically generating a fault diagnosis report and pushing it to a maintenance terminal.

[0007] As a preferred solution of the intelligent analysis method for monitoring distribution station area equipment described in the present invention, the following steps are included: using a deep learning model to extract features from real-time operation data to obtain an operation status feature vector: constructing a multi-layer convolutional neural network model, the multi-layer convolutional neural network model includes an input layer, multiple convolutional layers, a pooling layer and a fully connected layer; organizing the real-time operation data into a three-dimensional tensor as the input of the multi-layer convolutional neural network model; extracting features from the three-dimensional tensor through the convolution layer to generate a feature map; using the pooling layer to reduce the dimension of the feature map to retain the main feature information; repeating the feature extraction and dimensionality reduction process to form a multi-layer convolution and pooling structure to gradually extract high-level features; mapping the output of the last pooling layer to the vector space through the fully connected layer; normalizing the output of the fully connected layer to obtain an operation status feature vector.

[0008] As a preferred solution of the intelligent analysis method for monitoring distribution station area equipment described in the present invention, wherein: feature extraction of a three-dimensional tensor through a convolution layer includes setting multiple convolution kernels according to the size of the input three-dimensional tensor; executing a convolution feature extraction process: sliding the convolution kernel on the input three-dimensional tensor with a step size of n, and performing a convolution operation on each position; for each convolution position, calculating the element-wise product sum of the convolution kernel and the current data window to obtain a convolution result; and performing a nonlinear transformation on the convolution result through a ReLU activation function.

[0009] As a preferred solution of the intelligent analysis method for monitoring equipment in distribution station area described in the present invention, obtaining a preliminary abnormality degree quantitative index includes the following steps: constructing a KD tree index of a normal operating state feature vector; based on the KD tree index, calculating the Euclidean distance between the current operating state feature vector and the K nearest neighbor normal operating state feature vectors; inputting the K Euclidean distances into a Gaussian kernel function to obtain a nonlinear transformation result; normalizing the nonlinear transformation result to obtain a similarity value; determining an initial similarity threshold; dynamically adjusting the initial similarity threshold to obtain a second similarity threshold; comparing the similarity value and the second similarity threshold, and outputting a preliminary abnormality degree quantitative index X.

[0010] As a preferred solution of the intelligent analysis method for monitoring equipment in a distribution area of ​​the present invention, the following steps are included in the equipment status evaluation and classification based on the preliminary abnormality degree quantitative index: determining the multi-level classification threshold T 1 、T 2 、T 3 ; Compare the preliminary abnormality degree quantitative index X with the multi-level classification threshold to determine the preliminary state category; set a fuzzy interval [T i -δ,T i +δ], where δ represents the width; when the preliminary abnormality degree quantitative index X falls into the fuzzy interval [T i-δ,T i +δ], calculate the membership of X to two adjacent state categories; construct a comprehensive scoring function; based on the comprehensive scoring function, determine the equipment state evaluation result; the equipment state evaluation result includes the final state category of the equipment and the classification reliability R.

[0011] As a preferred solution of the intelligent analysis method for monitoring equipment in a distribution area of ​​the present invention, the specific formula of the comprehensive scoring function is as follows:

[0012] S=w 1 *X+w 2 *|μ_upper(X)-μ_lower(X)|+w 3 *E

[0013] Among them, X represents the preliminary abnormality degree quantitative index, |μ_upper(X)-μ_lower(X)| represents the absolute value of the membership difference, E represents the expert evaluation score, and w 1 、w 2 、w 3 Both represent weight coefficients, and w 1 +w 2 +w 3 .

[0014] As a preferred solution of the intelligent analysis method for monitoring equipment in a distribution station area described in the present invention, the following steps are included: automatically generating a fault diagnosis report and pushing it to a maintenance terminal: parsing the equipment status evaluation results, extracting the equipment status category and classification reliability R; matching the processing strategy from a preset strategy library according to the equipment status category; collecting diagnostic information based on the processing strategy, and generating a diagnostic report using an adaptive template; assigning a push priority to the diagnostic report according to the equipment status category and classification reliability R; pushing the diagnostic report to a designated maintenance terminal in the order of push priority; the calculation formula for the push priority is as follows:

[0015]

[0016] Among them, Severity is the severity of the state, R is the classification reliability, and Base priority is the base priority score.

[0017] In the second aspect, an embodiment of the present invention provides an intelligent analysis method system for monitoring equipment in a distribution station area, which includes a data acquisition and preprocessing module for acquiring and preprocessing real-time operating data of equipment in a distribution station area; a feature extraction module for using a deep learning model to perform feature extraction on real-time operating data to obtain an operating state feature vector; an abnormality degree quantification module for calculating the similarity between the operating state feature vector and the normal operating state feature vector in a historical database to obtain a preliminary abnormality degree quantification index; a state assessment and classification module for evaluating and classifying equipment states according to the preliminary abnormality degree quantification index to obtain equipment state assessment results; a diagnostic report generation and push module for taking corresponding measures according to the equipment state assessment results, automatically generating a fault diagnosis report and pushing it to a maintenance terminal.

[0018] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the intelligent analysis method for monitoring distribution station area equipment as described in the first aspect of the present invention are implemented.

[0019] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the intelligent analysis method for monitoring equipment in a distribution station area as described in the first aspect of the present invention are implemented.

[0020] The beneficial effects of the present invention are as follows: First, the multi-parameter monitoring and deep learning feature extraction technology adopted by the present invention improves the accuracy and comprehensiveness of anomaly detection, and can identify potential fault features that are difficult to detect with some traditional methods. Secondly, the introduced fuzzy interval and comprehensive scoring mechanism improves the accuracy of state assessment, helps to deal with boundary conditions, and reduces the possibility of misjudgment. Thirdly, the adaptive push mechanism designed by the present invention adjusts the priority according to the severity of the state and the reliability of classification, which is conducive to the rational allocation of maintenance resources and improves operation and maintenance efficiency. Finally, by automatically generating diagnostic reports and combining the human-computer collaboration mechanism, the system provides effective decision-making support for maintenance personnel, while retaining the necessary space for manual intervention, realizing the combination of intelligence and professional judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a framework flow chart of the intelligent analysis method for monitoring equipment in distribution station areas.

[0023] Figure 2 Feature extraction process for intelligent analysis method of equipment monitoring in distribution station area.

[0024] Figure 3 A flow chart for generating and pushing fault diagnosis reports for the intelligent analysis method for monitoring equipment in distribution station areas.

[0025] Figure 4 This is a module flow chart of the intelligent analysis method for monitoring equipment in distribution station areas.

[0026] Figure 5 Computer equipment diagram for intelligent analysis method of equipment monitoring in distribution station area. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0030] Example 1

[0031] Reference Figure 1 to Figure 5 , which is the first embodiment of the present invention, and provides an intelligent analysis method for monitoring equipment in a distribution station area, including:

[0032] S1: Collect and pre-process the real-time operation data of the equipment in the distribution area.

[0033] It should be noted that many current distribution substation equipment monitoring systems have the following problems in data collection: the collection parameters are not comprehensive, often focusing only on basic parameters such as voltage and current; the sampling frequency is low, making it difficult to capture instantaneous anomalies; the data preprocessing capability is limited and easily affected by noise. In response to the above problems, the present invention selects voltage, current, power factor, and temperature as monitoring parameters. The selection of these parameters is based on the following considerations: voltage indicates power supply quality and system stability, current reflects load conditions, power factor reflects energy utilization efficiency, and temperature is directly related to equipment life and safe operation.

[0034] Preferably, through multi-parameter monitoring, a comprehensive assessment of the equipment status can be achieved, abnormalities can be discovered in time, and failures can be prevented, thereby improving power supply reliability and equipment management efficiency.

[0035] S2: Use the deep learning model to extract features from real-time operation data and obtain the operation status feature vector.

[0036] It should be noted that, considering the characteristics of high dimensionality, nonlinearity, and strong time-series correlation of equipment operation data in distribution substations, traditional feature extraction methods are difficult to effectively capture complex data patterns and potential fault characteristics. The present invention adopts a multi-layer convolutional neural network model for feature extraction. This method has the following advantages: it can automatically learn and extract complex nonlinear features; it is suitable for processing multidimensional data with spatiotemporal structures; it has good generalization ability and can adapt to different types of equipment and working conditions.

[0037] Specifically, the following steps are included:

[0038] S2.1: Build a multi-layer convolutional neural network model, including input layer, convolution layer, pooling layer and fully connected layer.

[0039] Preferably, this multi-layer structure can gradually extract features from low-level to high-level, such as Figure 2 As shown, complex patterns in the data can be effectively captured.

[0040] S2.2: Organize the real-time running data into a three-dimensional tensor according to time sequence and parameter type as the input of the multi-layer convolutional neural network model.

[0041] S2.3: Extract features from the three-dimensional tensor through the convolution layer and generate a feature map.

[0042] Preferably, according to the size of the input three-dimensional tensor, multiple 3×3×C convolution kernels are set, where C is the number of channels of the input data; executing the convolution feature extraction process includes: sliding the convolution kernel on the input three-dimensional tensor with a step size of n, and performing a convolution operation on each position; for each convolution position, calculating the element-wise product sum of the convolution kernel and the current data window to obtain a convolution result; performing a nonlinear transformation on the convolution result through a ReLU activation function; repeating the convolution feature extraction process for each convolution kernel, each convolution kernel generates a two-dimensional feature map, thereby obtaining multiple feature maps; stacking all feature maps to form a new three-dimensional feature tensor.

[0043] It should be noted that the purpose of generating a new three-dimensional feature tensor is to maintain the structured representation of the data for subsequent processing. This method stacks multiple two-dimensional feature maps into a three-dimensional tensor, where the third dimension represents the features extracted by different convolution kernels. This structure not only retains the spatial relationship of the original data, but also extracts a higher-level feature representation through convolution operations. It enables subsequent pooling layers or further convolution operations to perform a comprehensive analysis of different features while retaining the spatial relationship, thereby providing a rich information basis for condition monitoring and fault diagnosis.

[0044] S2.4: Use the maximum pooling layer to reduce the dimension of the feature map and retain the main feature information.

[0045] Preferably, a maximum pooling layer is introduced in the convolutional neural network for dimensionality reduction operation, mainly based on the following considerations: reducing computational complexity; the maximum pooling operation retains the maximum value in the local area, which usually corresponds to the most significant feature and helps the model focus on important information; dimensionality reduction reduces the number of model parameters, suppresses overfitting to a certain extent, and improves the generalization ability of the model; combining convolution kernels of different sizes and pooling operations can achieve multi-scale feature extraction and enhance the model's recognition ability for patterns of different sizes.

[0046] S2.5: Repeat the feature extraction and dimensionality reduction process to form a multi-layer convolution and pooling structure to gradually extract high-level features.

[0047] S2.6: Map the output of the last pooling layer to a vector space of fixed dimension through a fully connected layer.

[0048] S2.7: Perform L2 normalization on the output of the fully connected layer to obtain a running status feature vector of unit length.

[0049] S3: Calculate the similarity between the operating status feature vector and the normal operating status feature vector in the historical database to obtain a quantitative index of the abnormality degree.

[0050] Specifically, a KD tree algorithm is used to construct a KD tree index of the normal operating status feature vector in the historical database; based on the KD tree index, a K nearest neighbor algorithm is used to calculate the Euclidean distance between the current operating status feature vector and the K nearest neighbor normal operating status feature vectors; the calculated K Euclidean distances are input into a preset Gaussian kernel function to obtain a nonlinear transformation result; the nonlinear transformation result is normalized to obtain a similarity value; based on the statistical analysis results of the historical data, an initial similarity threshold for distinguishing between normal and abnormal states is determined; according to the equipment operation data and expert evaluation results within a set period, the initial similarity threshold is dynamically adjusted to obtain a second similarity threshold; the similarity value and the second similarity threshold are compared to output a preliminary abnormality degree quantitative index X.

[0051] S4: Evaluate and classify the equipment status according to the preliminary abnormality degree quantitative index X to obtain the equipment status evaluation result.

[0052] Specifically, the following steps are included:

[0053] S4.1: Determine the multi-class classification threshold T 1 、T 2 、T 3 .

[0054] It should be noted that the multi-class classification threshold T 1 、T 2 、T 3 Determined based on historical data and expert experience, and T 1 Indicates the boundary between normal and slightly abnormal, T 2 Indicates the boundary between mild abnormality and moderate abnormality, T 3 Indicates the boundary between moderate abnormality and severe abnormality.

[0055] S4.2: Compare the preliminary abnormality degree quantitative index X with the multi-level classification threshold to determine the preliminary state category.

[0056] Specifically, when X<T 1 When T 1 ≤X<T 2 When T 2 ≤X<T 3 When X>T 3 , the preliminary status category is severe abnormality.

[0057] S4.3: Set a fuzzy interval with a width of δ on both sides of each classification threshold [T i -δ,T i +δ], where i = 1, 2, 3.

[0058] It should be noted that the δ value is determined by analyzing historical data. Specifically: calculate the standard deviation δ of each category i ; Set δ = min(δ i )*k, where k is an adjustable parameter (the default value can be set to 1.5).

[0059] Preferably, this method is data-driven, has a statistical basis, provides flexibility and adaptability, and can adjust the fuzzy interval width according to the actual data distribution characteristics. At the same time, by introducing an adjustable parameter k, it allows fine-tuning according to specific application requirements or expert knowledge.

[0060] S4.4: When the preliminary abnormality quantitative indicator X falls into the fuzzy interval [T i-δ,T i +δ], the membership of X to two adjacent state categories is calculated.

[0061] It should be noted that the membership of X to two adjacent state categories is calculated as follows:

[0062]

[0063] in,

[0064] μ_lower(X)+μ_upper(X)=1

[0065] Among them, X represents the quantitative index of the initial abnormality degree, T i represents the i-th multi-level classification threshold, δ represents the half width of the fuzzy interval (the full width of the fuzzy interval is 2δ); μ_lower(X) represents the effect of X on the current classification threshold T i The membership of the lower state category, μ_upper(X) represents the degree of membership of X to the current classification threshold T i The degree of membership in the upper state category.

[0066] It should be noted that the value range of μ_lower(X) and μ_upper(X) is [0,1]; and the boundary value is processed as follows: when X = T i -δ, μ_lower(X)=1, μ_upper(X)=0; when X=T i When +δ, μ_lower(X)=0, μ_upper(X)=1.

[0067] For example, suppose T 2 =0.5 (the boundary between slight anomaly and moderate anomaly), δ = 0.1, then the fuzzy interval is [0.4, 0.6]. If X = 0.45, then μ_lower(X) = 0.75, μ_upper(X) = 0.25; this means that X = 0.45 has a membership of 0.75 for slight anomaly and 0.25 for moderate anomaly.

[0068] S4.5: Construct a comprehensive scoring function.

[0069] Specifically, the specific formula of the comprehensive scoring function is as follows:

[0070] S=w 1 *X+w 2 *|μ_upper(X)-μ_lower(X)|+w 3 *E

[0071] Among them, X represents the preliminary abnormality degree quantitative index, |μ_upper(X)-μ_lower(X)| represents the absolute value of the membership difference, E represents the expert evaluation score, and w 1 、w 2 、w 3 Both represent weight coefficients, and w 1 +w 2 +w 3 .

[0072] It should be noted that if the preliminary abnormality quantitative index X does not fall into the fuzzy interval [T i -δ,T i +δ], then w 2 =0, and the expert evaluation score range is also [0,1].

[0073] S4.6: Based on the comprehensive scoring function, determine the equipment status assessment results including the equipment final status category and classification reliability.

[0074] Specifically, the comprehensive score S is calculated; the comprehensive score S is compared with the multi-level classification threshold to determine the final state category of the equipment; the classification reliability R is calculated, and if the preliminary abnormality degree quantitative index X falls into the fuzzy interval [T i -δ,T i +δ], the formula is as follows:

[0075] R=1-|μ_upper(X)-μ_lower(X)|

[0076] Among them, μ_lower(X) represents the value of X for the current classification threshold T i The membership of the lower state category, μ_upper(X) represents the degree of membership of X to the current classification threshold T i The degree of membership in the upper state category.

[0077] If the preliminary abnormality quantitative index X does not fall into the fuzzy interval [T i -δ,T i +δ], then R=1.

[0078] It should be noted that the value range of classification reliability R is from 0 to 1, and the larger the value, the more reliable the classification result.

[0079] S5: Take corresponding measures according to the equipment status assessment results, automatically generate a fault diagnosis report and push it to the maintenance terminal.

[0080] Specifically, parse the equipment status assessment results, extract the equipment status category and classification reliability R; match the processing strategy from the preset strategy library according to the equipment status category; collect diagnostic information based on the processing strategy, and use an adaptive template to generate a diagnostic report, including fault description, potential causes, recommended measures, and reliability description (when R is low, the reminder requires further manual confirmation); assign push priority to the diagnostic report according to the equipment status category and classification reliability; push the diagnostic report to the designated maintenance terminal in the order of push priority.

[0081] It should be noted that the processing strategies are matched from the preset strategy library, as follows: normal status matches the daily monitoring strategy; minor abnormalities match the preventive maintenance strategy; moderate abnormalities match the priority maintenance strategy; severe abnormalities match the emergency processing strategy.

[0082] Furthermore, assigning push priority to the diagnostic report includes the following steps: defining the state severity (normal = 1, slightly abnormal = 2, moderately abnormal = 3, severely abnormal = 4); calculating the basic priority score, the specific formula is as follows:

[0083] Base priority =R*Severity

[0084] Among them, Severity is the severity of the state and R is the classification reliability.

[0085] Furthermore, the push priority score is calculated, and the specific formula is as follows:

[0086]

[0087] Among them, Severity is the severity of the state, R is the classification reliability, and Base priority is the base priority score.

[0088] Preferably, by integrating equipment status evaluation, adaptive processing strategy and intelligent push mechanism, the present invention constructs an efficient, flexible and reliable equipment management system. The system can not only accurately identify and classify abnormal equipment status, but also dynamically adjust the processing priority according to the classification reliability, thereby optimizing resource allocation. The automatically generated diagnostic report provides comprehensive decision support for maintenance personnel, while the human-machine collaboration mechanism ensures the reliability of system judgment. In addition, the application of preventive maintenance strategies helps to extend equipment life and reduce maintenance costs.

[0089] Furthermore, the present embodiment also provides a distribution substation area equipment monitoring intelligent analysis method system, including a data acquisition and preprocessing module, which is used to acquire and preprocess the real-time operating data of the distribution substation area equipment; a feature extraction module, which is used to extract features from the real-time operating data using a deep learning model to obtain an operating state feature vector; an abnormality degree quantification module, which is used to calculate the similarity between the operating state feature vector and the normal operating state feature vector in the historical database to obtain a preliminary abnormality degree quantification index; a state assessment and classification module, which is used to assess and classify the equipment state according to the preliminary abnormality degree quantification index to obtain an equipment state assessment result; and a diagnostic report generation and push module, which is used to take corresponding measures according to the equipment state assessment result, automatically generate a fault diagnosis report, and push it to the maintenance terminal. Specifically, Figure 4 shown.

[0090] This embodiment also provides a computer device, which is suitable for the case of an intelligent analysis method for monitoring equipment in a distribution station area, and includes a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the intelligent analysis method for monitoring equipment in a distribution station area as proposed in the above embodiment.

[0091] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc. Specifically, Figure 5 shown.

[0092] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for intelligent analysis of equipment monitoring in a distribution station area as proposed in the above embodiment is implemented.

[0093] First, the multi-parameter monitoring and deep learning feature extraction technology adopted by the present invention improves the accuracy and comprehensiveness of anomaly detection, and can identify potential fault features that are difficult to detect with some traditional methods. Secondly, the introduced fuzzy interval and comprehensive scoring mechanism improves the accuracy of state assessment, helps to deal with boundary conditions, and reduces the possibility of misjudgment. Third, the adaptive push mechanism designed by the present invention adjusts the priority according to the severity of the state and the reliability of classification, which is conducive to the rational allocation of maintenance resources and improves operation and maintenance efficiency. Finally, by automatically generating diagnostic reports and combining the human-computer collaboration mechanism, the system provides effective decision-making support for maintenance personnel, while retaining the necessary space for manual intervention, realizing the combination of intelligence and professional judgment.

[0094] Example 2

[0095] Reference Figure 1 to Figure 5 , which is the second embodiment of the present invention, provides an intelligent analysis method for monitoring equipment in a distribution station area. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0096] A six-month trial run of the intelligent monitoring system was conducted in a typical urban distribution substation in a provincial capital city in East China. The distribution substation is located in the downtown commercial district, covers an area of ​​about 0.8 square kilometers, includes one S11-M-1000 / 10 capacity transformer and four 10kV outgoing lines, and serves about 180 residential users, 15 commercial complexes and two schools.

[0097] During the trial operation, the system collected key parameters such as voltage, current, power factor and temperature of the transformer and each outgoing line. The sampling frequency was increased to once every 2 seconds, which is a significant improvement compared to the original sampling method of once every 3 minutes. During the 180-day trial operation, the system collected a total of about 77.76 million valid data points. Through the improved data preprocessing algorithm, 2.8% of abnormal data was successfully eliminated, laying a solid foundation for subsequent analysis.

[0098] An improved deep learning model was used in the feature extraction stage. The model contains 4 convolutional layers and 3 pooling layers. The input is the data of a 24-hour sliding window (43,200 data points), and the output is a 192-dimensional feature vector. The model training used nearly 3 years of historical operation data, and the accuracy on the test set reached 94.2%. In actual application, the system generates a feature vector every 30 minutes, and a total of 8,640 operation status feature vectors were generated in half a year.

[0099] A reference database containing 25,000 samples of normal operating status was constructed during the anomaly detection phase. The system uses an improved K-nearest neighbor algorithm (K=7) for similarity calculation. The initial similarity threshold was set to 0.878 through data analysis, and after 5 fine-tunings during the trial run, it finally stabilized at 0.861. The data showed that about 1.9% of the time periods were marked as potential abnormal states, of which 0.3% were confirmed as real anomalies.

[0100] A multi-level fuzzy classification method was used in the status assessment stage. The cut-off values ​​for normal-slightly abnormal, slightly-moderately abnormal, and moderately-severely abnormal were 0.83, 0.69, and 0.55, respectively. The range of ±0.04 on both sides of each cut-off value was defined as a fuzzy interval. During the six-month trial run, about 1.7% of the data fell into the fuzzy interval. By introducing expert experience weights (values ​​of 0.65, 0.25, and 0.1), the final status classification accuracy of the system reached 97.3%, of which high-confidence (>0.92) classification results accounted for 93.5%.

[0101] Based on the evaluation results, the system automatically matches the corresponding processing strategy and generates a diagnostic report. During the trial operation, a total of 246 reports were generated, including 185 daily monitoring reports, 42 preventive maintenance recommendations, 16 priority maintenance notices, and 3 emergency processing alarms. The intelligent push algorithm ensures that 95% of medium and high priority reports are delivered to the terminals of relevant personnel within 3 minutes. Through timely preventive maintenance, two potential large-scale power outages were successfully avoided, and it is estimated that economic losses of about 800,000 yuan were saved.

[0102] During the entire trial operation, the system's average response time was 0.8 seconds, far superior to traditional manual monitoring methods. The work efficiency of maintenance personnel increased by about 25%, and the false alarm rate was reduced from the original 4.2% to 1.4%. In addition, the system's predictive maintenance recommendations helped extend the expected life of the transformer by about 8% and reduced annual maintenance costs by 17.3%.

[0103] Preferably, the experimental results show that the comparison indicators of the present invention and the traditional method are as shown in Table 1.

[0104] Table 1 Comparison indexes of the present invention and the traditional method

[0105] Comparison indicators Traditional methods The present invention Data sampling frequency 3 minutes / time 2 seconds / time Abnormal data elimination rate 1.2％ 2.8％ Feature extraction model accuracy 92.8％ 97.3％ Proportion of high-confidence classification results 85.2％ 93.5％ Status classification accuracy 87.8％ 92.5％ Average system response time 3.5s 0.8s Improved work efficiency of maintenance personnel - 25％ False Positive Rate 4.2％ 1.4％

[0106] As can be seen from Table 1, the present invention is significantly superior to traditional methods in multiple key indicators. In particular, the new system shows obvious advantages in terms of data sampling frequency, anomaly detection capability, classification accuracy and system response time. In addition, the new system also brings about improved maintenance efficiency, reduced false alarm rate, extended equipment life and reduced maintenance costs, which are difficult to achieve with traditional methods. In general, the intelligent monitoring system of the present invention has significant technical advantages and economic benefits in improving the operational reliability and management efficiency of distribution station areas.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Intelligent analysis method for monitoring equipment in distribution area, characterized by: include, Collect and pre-process the real-time operation data of the equipment in the distribution area; The real-time operation data includes voltage, current, power factor, and temperature parameters; Using a deep learning model to extract features from the real-time operation data to obtain an operation status feature vector; Calculate the similarity between the operating state feature vector and the normal operating state feature vector in the historical database to obtain a preliminary quantitative index of abnormality degree; Evaluate and classify the equipment status based on the preliminary abnormality degree quantitative indicators to obtain the equipment status evaluation results; According to the equipment status evaluation results, corresponding measures are taken to automatically generate a fault diagnosis report and push it to the maintenance terminal.

2. The intelligent analysis method for monitoring equipment in a distribution area according to claim 1, characterized in that: The use of a deep learning model to extract features from the real-time operation data to obtain an operation status feature vector comprises the following steps: Constructing a multi-layer convolutional neural network model, wherein the multi-layer convolutional neural network model includes an input layer, multiple convolutional layers, a pooling layer, and a fully connected layer; Organizing the real-time running data into a three-dimensional tensor as an input of the multi-layer convolutional neural network model; Performing feature extraction on the three-dimensional tensor through the convolution layer to generate a feature map; Using the pooling layer to reduce the dimension of the feature map and retain the main feature information; Repeat the feature extraction and dimensionality reduction process to form a multi-layer convolution and pooling structure to gradually extract high-level features; Mapping the output of the last pooling layer to the vector space through the fully connected layer; The output of the fully connected layer is normalized to obtain a running state feature vector.

3. The intelligent analysis method for monitoring equipment in a distribution area according to claim 2, characterized in that: The extracting features of the three-dimensional tensor by the convolution layer includes: Set multiple convolution kernels according to the size of the input three-dimensional tensor; Execute the convolution feature extraction process: Slide the convolution kernel on the input three-dimensional tensor with a step size n, and perform a convolution operation on each position; For each convolution position, the element-wise product sum of the convolution kernel and the current data window is calculated to obtain a convolution result; The convolution result is nonlinearly transformed by the ReLU activation function.

4. The intelligent analysis method for monitoring equipment in a distribution area according to claim 1, characterized in that: The obtaining of the preliminary abnormality degree quantitative index comprises the following steps: Construct a KD tree index of the normal operation status feature vector; Based on the KD tree index, calculating the Euclidean distance between the current running state feature vector and the K nearest neighbor normal running state feature vectors; Input K Euclidean distances into the Gaussian kernel function to obtain the nonlinear transformation result; Normalizing the nonlinear transformation result to obtain a similarity value; Determine an initial similarity threshold; Dynamically adjusting the initial similarity threshold to obtain a second similarity threshold; Compare the similarity value with the second similarity threshold, and output a preliminary abnormality degree quantitative index X.

5. The intelligent analysis method for monitoring equipment in a distribution area according to claim 4, characterized in that: The device status evaluation and classification according to the preliminary abnormality degree quantitative indicators includes the following steps: Determine the multi-class classification thresholds T1, T2, and T3; Compare the preliminary abnormality degree quantitative index X with the multi-level classification threshold to determine the preliminary state category; A fuzzy interval [T i -δ,T i +δ], where δ represents the width; When the preliminary abnormality degree quantitative indicator X falls into the fuzzy interval [T i -δ,T i +δ], calculate the membership of X to two adjacent state categories; Construct a comprehensive scoring function; Based on the comprehensive scoring function, an equipment state evaluation result is determined; the equipment state evaluation result includes a final state category of the equipment and a classification reliability R.

6. The intelligent analysis method for monitoring equipment in a distribution area according to claim 5, characterized in that: The specific formula of the comprehensive scoring function is as follows: S=w1*X+w2*|μ_upper(X)-μ_lower(X)|+w3*E Among them, X represents the preliminary abnormality degree quantitative index, |μ_upper(X)-μ_lower(X)| represents the absolute value of the membership difference, E represents the expert evaluation score, w1, w2, and w3 all represent weight coefficients, and w1+w2+w3.

7. The intelligent analysis method for monitoring equipment in a distribution area according to claim 5, characterized in that: The automatic generation of a fault diagnosis report and pushing it to a maintenance terminal comprises the following steps: Analyze the equipment status assessment results, extract the equipment status category and classification reliability R; According to the device status category, matching a processing strategy from a preset strategy library; Collecting diagnostic information based on the processing strategy and generating a diagnostic report using an adaptive template; According to the device status category and the classification reliability R, a push priority is assigned to the diagnostic report; and the diagnostic report is pushed to a designated maintenance terminal in the order of push priority; The calculation formula of the push priority is as follows: Among them, Severity is the severity of the state, R is the classification reliability, and Base priority is the base priority score.

8. A distribution substation area equipment monitoring intelligent analysis method system, based on the distribution substation area equipment monitoring intelligent analysis method according to any one of claims 1 to 7, characterized in that: Also includes, Data collection and preprocessing module, used to collect and preprocess the real-time operation data of the equipment in the distribution area; A feature extraction module, used to extract features from the real-time operation data using a deep learning model to obtain an operation state feature vector; An abnormality degree quantification module, used to calculate the similarity between the operating state feature vector and the normal operating state feature vector in the historical database to obtain a preliminary abnormality degree quantification index; The status assessment and classification module is used to assess and classify the equipment status according to the preliminary abnormality degree quantitative indicators to obtain the equipment status assessment results; The diagnostic report generation and push module is used to take corresponding measures according to the equipment status evaluation results, automatically generate a fault diagnosis report and push it to the maintenance terminal.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent analysis method for monitoring equipment in a distribution station area described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent analysis method for monitoring equipment in a distribution station area described in any one of claims 1 to 7 are implemented.

Images