Intelligent state identification regulation and control method and system for electrical equipment

Abstract

The invention relates to the technical field of electrical equipment monitoring, and particularly discloses an electrical equipment intelligent state identification regulation and control method and system, equipment operation data is collected in real time through multiple sensors such as current and voltage sensors, historical and power grid load information is combined, after filtering and normalization preprocessing, features are extracted by using a convolutional neural network, and the state of the electrical equipment is identified. Inputting a support vector machine model to identify normal, slight abnormal and other states of the equipment; according to a state result, a regulation and control strategy is generated in combination with a power grid load and operation parameters, parameters are regulated during slight abnormity, power failure is given during serious abnormity or faults, and real-time monitoring, feedback and optimization are carried out. Through multi-sensor data acquisition, an advanced data processing algorithm and an intelligent regulation and control strategy, accurate identification and effective regulation and control of the operation state of the electrical equipment are realized, the operation reliability and stability of the electrical equipment are improved, and safe and efficient operation of a power system is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment monitoring, in particular to an intelligent state recognition and regulation method and system for electrical equipment. BACKGROUND

[0002] With the continuous development of power systems and the increasing demand for intelligentization, the stable operation of electrical equipment is crucial for ensuring the reliability and safety of power supply. Traditional methods of electrical equipment state monitoring mainly rely on manual inspection and simple instrument measurement, which has low monitoring efficiency, poor real-time performance, and difficulty in discovering potential faults. Although some existing automatic monitoring technologies can collect part of the equipment operation data, there is still room for improvement in data processing, state recognition accuracy, and intelligent regulation strategies. For example, the feature extraction of complex operating states of electrical equipment is not comprehensive enough, resulting in low accuracy of state recognition. When the equipment is abnormal, the regulation strategy lacks flexibility and adaptability, and cannot be effectively adjusted according to the actual situation of the equipment and the power grid load.

[0003] Therefore, an intelligent state recognition and regulation method and system for electrical equipment have become a problem that needs to be solved. SUMMARY

[0004] The technical problem to be solved by the present application is to provide an intelligent state recognition and regulation method and system for electrical equipment, which realizes accurate recognition and effective regulation of the operating state of electrical equipment through multi-sensor data acquisition, advanced data processing algorithms and intelligent regulation strategies, improves the operating reliability and stability of electrical equipment, and ensures the safe and efficient operation of the power system.

[0005] To solve the above technical problems, the technical solution provided by the present application is as follows: an intelligent state recognition and regulation method for electrical equipment, comprising the following steps:

[0006] S1, using current sensors, voltage sensors, temperature sensors, vibration sensors and humidity sensors, real-time acquisition of current, voltage, temperature, vibration frequency and environmental humidity data of electrical equipment operation, and simultaneously obtaining device historical operation data, maintenance records and power grid load information through a network communication module;

[0007] S2, filtering the collected raw data to remove noise, and using a normalization method to standardize different types and magnitudes of data;

[0008] S3, using a convolutional neural network to extract features from the preprocessed data to obtain a key feature vector reflecting the operating state of the electrical equipment;

[0009] S4, input the extracted feature vector into a state recognition model constructed based on a support vector machine and trained by a large amount of historical data, to determine that the electrical equipment is in a normal operation, slight abnormality, serious abnormality or fault state;

[0010] S5, according to the equipment state recognition result, combining the power grid load information and the equipment operation parameters, a control strategy is generated: when the equipment is in slight abnormality, the operation parameters are adjusted, when it is in serious abnormality or fault, an alarm is issued and the power supply is cut off according to the preset protection logic;

[0011] S6, the control strategy is sent to the electrical equipment actuator through the control module, and the control effect is monitored in real time and feedback is optimized.

[0012] Further, the preprocessed multi-dimensional time series data is divided into a three-dimensional tensor input according to a time window, and the tensor dimension is [time step × sensor type number × feature dimension];

[0013] A multi-layer convolutional neural network architecture is constructed, which includes in turn:

[0014] The first one-dimensional convolutional layer adopts 16 convolutional kernels with a size of 3 and a step of 1, and the activation function is ReLU, which is used to extract short-term time series features;

[0015] The maximum pooling layer has a pooling window size of 2 and a step of 2, which is used for feature dimension reduction and noise suppression;

[0016] The second one-dimensional convolutional layer adopts 32 convolutional kernels with a size of 5 and a step of 1, and the activation function is LeakyReLU, which is used to extract medium-term time series dependence;

[0017] The global average pooling layer compresses the time series dimension into a feature vector;

[0018] Batch normalization layers and Dropout layers are inserted between the convolutional layers, and the Dropout rate is 0.2, which is used to prevent overfitting and accelerate training;

[0019] The feature is mapped to a low-dimensional space through a fully connected layer to generate a key feature vector with a dimension of 64, which contains time-frequency domain features, multi-sensor correlation features and environmental influence features of the electrical equipment operation state;

[0020] The network is trained using a triplet loss function to enhance the aggregation of similar state features and the separation of different state features, and the training data includes normal, abnormal and fault state samples.

[0021] Further, the specific method for determining the electrical equipment operation state is as follows:

[0022] The 64-dimensional key feature vector extracted by the convolutional neural network is taken as input, and the normal, slight abnormal, serious abnormal and fault four states of the corresponding device are taken as output labels;

[0023] An SVM-RBF kernel function is used to construct a multi-class state recognition model, and an One-vs-One strategy is used to convert the four-class problem into six two-class sub-problems, each of which is responsible for distinguishing two states;

[0024] The key feature vector is subjected to Z-score standardization, and the Mahalanobis distance of the key feature vector to the center of the four states is calculated; the final state category is determined through a voting mechanism: six binary classifiers are used for voting, if the number of votes of a certain state exceeds the threshold value, and the Mahalanobis distance of the sample to the state is the smallest, then it is determined that the device is in the state; when the difference between the votes of any two states is less than 5%, the fuzzy state determination is triggered;

[0025] The method for determining the fuzzy state is: calculating the distance of the key feature vector to each classification hyperplane, if all are greater than the threshold value, then it is determined as unknown state, otherwise, the two closest categories are selected, and the historical state transition probability of the device is combined for secondary determination;

[0026] When 5 consecutive samples are determined as unknown states, the incremental learning process is started; the most representative 20 samples are selected for manual labeling through active learning, and the training set is updated; a sliding window mechanism is used, and every time 100 labeled samples are added, the earliest 50 samples are discarded, so that the size of the training set is stable.

[0027] Further, in the S5 step, when the device is in a slight abnormal state, the specific way of adjusting the operating parameters is: according to the parameter distribution of the device in the normal state in the historical operation data, combined with the current power grid load information, the voltage, current and power parameters of the electrical equipment are adjusted through the preset parameter adjustment rule, and the adjusted device operation data is recorded for subsequent model optimization.

[0028] Further, in the S6 step, when the real-time monitoring and control effect is monitored, the effectiveness of the control strategy is judged by comparing the change amplitude of the electrical equipment operating parameters before and after the control, and whether the state recognition result returns to normal; if the control effect does not reach the expectation, the control strategy is adjusted according to the preset feedback optimization rule and executed again.

[0029] The application also provides an intelligent state recognition and control system of electrical equipment, which is used for realizing the above method, comprising:

[0030] The data acquisition module comprises a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, a humidity sensor and a network communication module, and is used for acquiring real-time electrical equipment operation data, historical data, maintenance records and power grid load information;

[0031] Data preprocessing module: filtering and normalizing the collected raw data;

[0032] Feature extraction module: based on convolutional neural network, the preprocessed data is feature extracted to generate key feature vector;

[0033] State recognition module: using support vector machine model, the running state of electrical equipment is judged according to the extracted key feature vector;

[0034] Control strategy generation module: according to the device state, power grid load and operation parameters, the corresponding control strategy is generated;

[0035] Control execution and feedback module: the control strategy is sent to the execution mechanism, and the control effect is monitored in real time, and the optimization is carried out according to the feedback.

[0036] Compared with the prior art, the advantages of the present application are:

[0037] The present application can fully and accurately mine the running state characteristics of electrical equipment by collecting rich data through multiple types of sensors, combining historical and power grid load information, and using advanced data preprocessing and feature extraction algorithms, and improve the accuracy of state recognition.

[0038] The state recognition model based on support vector machine of the present application can effectively handle complex state determination conditions by combining multiple determination strategies and incremental learning mechanism, and improve the reliability and adaptability of state recognition.

[0039] The present application can effectively guarantee the safe and stable operation of electrical equipment and improve the reliability and efficiency of power system by generating reasonable control strategy according to different states of equipment and power grid conditions, and continuously improving the control effect through feedback optimization mechanism.

[0040] The intelligent state recognition and control system matched by the present application, the modules work collaboratively, realizes the full-process automation and intelligentization from data acquisition to control optimization, has good practicability and popularization value. BRIEF DESCRIPTION OF DRAWINGS

[0041] Fig. 1 It is a flow chart of the intelligent state recognition and control method of electrical equipment of the present application.

[0042] Fig. 2 It is a system block diagram of the intelligent state recognition and control system of electrical equipment of the present application. DETAILED DESCRIPTION

[0043] Various exemplary embodiments of the present application will be described herein below with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions, and numerical values set forth in these embodiments are not limiting to the scope of the present application unless specifically stated otherwise.

[0044] The following description of at least one exemplary embodiment is merely exemplary in nature and is in no way intended to limit the scope of the application its application or uses.

[0045] Techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail herein. However, where appropriate, such techniques, methods, and devices can be presumed to be a part of the specification.

[0046] In all of the examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not limiting. Thus, other examples of the exemplary embodiments can have different values.

[0047] The electrical equipment intelligent state recognition and regulation method and system will be further described in detail below in combination with the accompanying drawings.

[0048] In combination with the accompanying Figs. 1-2 The present application will be described in detail.

[0049] An electrical equipment intelligent state recognition and regulation method, specifically comprising the following steps:

[0050] Data acquisition: Use current sensors, voltage sensors, temperature sensors, vibration sensors, and humidity sensors to collect real-time current, voltage, temperature, vibration frequency, and environmental humidity data of electrical equipment operation. At the same time, obtain device historical operation data, maintenance records, and power grid load information through a network communication module. Multiple sensors work together to comprehensively collect equipment operation data, combined with historical and power grid load information, to provide a rich data foundation for subsequent analysis.

[0051] Data preprocessing: Filter the collected raw data to remove noise, and use normalization methods to standardize data of different types and magnitudes. Through filtering and normalization, improve data quality, make different types of data comparable, and facilitate subsequent algorithm processing.

[0052] Feature extraction: a convolutional neural network is used to extract features from the preprocessed data, obtaining a key feature vector reflecting the running state of the electrical equipment. Specifically, the preprocessed multi-dimensional time series data is divided into three-dimensional tensors according to the time window, and the tensor dimensions are [time step x sensor type number x feature dimension]; a multi-layer convolutional neural network architecture is constructed, which includes: a first one-dimensional convolutional layer with 16 convolutional kernels of size 3 and a step size of 1, and a ReLU activation function, used to extract short-term time series features; a max-pooling layer with a pooling window size of 2 and a step size of 2, used for feature dimension reduction and noise suppression; a second one-dimensional convolutional layer with 32 convolutional kernels of size 5 and a step size of 1, and a LeakyReLU activation function, used to extract medium-term time series dependencies; a global average pooling layer to compress the time series dimension into a feature vector; batch normalization layers and Dropout layers are inserted between the convolutional layers, with a Dropout rate of 0.2, used to prevent overfitting and accelerate training; a fully connected layer is used to map the features to a low-dimensional space, generating a key feature vector with a dimension of 64, which contains the time-frequency domain features, multi-sensor correlation features and environmental influence features of the electrical equipment running state; a triplet loss function is used to train the network to enhance the aggregation of similar state features and the separation of different state features, and the training data contains normal, abnormal and fault state samples. The convolutional neural network architecture can fully exploit the feature information in the data, improving the accuracy and effectiveness of feature extraction.

[0053] State recognition: input the extracted feature vector into the state recognition model built based on support vector machine and trained with a large amount of historical data to determine whether the electrical equipment is in normal operation, slight abnormality, serious abnormality or fault state. The specific determination method is: the 64-dimensional key feature vector extracted by the convolutional neural network is used as the input, and the normal, slight abnormality, serious abnormality and fault states of the equipment are used as the output labels; a multi-class state recognition model is built using SVM-RBF kernel function, and the One-vs-One strategy is used to convert the four-class problem into six two-class sub-problems, each of which is responsible for distinguishing between two states; the key feature vector is standardized by Z-score, and the Mahalanobis distance from the center of the four states is calculated; the final state category is determined by a voting mechanism: six binary classifiers are used for voting, if the number of votes for a certain state exceeds the threshold, and the Mahalanobis distance of the sample to the state is the smallest, then the equipment is determined to be in that state; when the difference between the votes of any two states is less than 5%, the fuzzy state determination is triggered; the method of fuzzy state determination is: calculate the distance from the key feature vector to each classification hyperplane, if they are all greater than the threshold, then it is determined as unknown state, otherwise, select the two closest categories, combine the historical state transition probability of the equipment for secondary determination; when 5 consecutive samples are determined as unknown state, start the incremental learning process; select the most representative 20 samples for manual annotation through active learning, update the training set; use a sliding window mechanism, discard the earliest 50 samples every time 100 labeled samples are added, keep the size of the training set stable. This state recognition method combines multiple strategies, improves the accuracy and reliability of state recognition, and can effectively handle complex state determination situations.

[0054] Control strategy generation: according to the state recognition result of the equipment, combined with the power grid load information and the equipment operation parameters, the control strategy is generated: when the equipment is in slight abnormality, adjust the operation parameters, when it is in serious abnormality or fault, issue an alarm and cut off the power supply according to the preset protection logic. When the equipment is in slight abnormality, the specific way to adjust the operation parameters is: according to the parameter distribution of the equipment in the normal state in the historical operation data, combined with the current power grid load information, adjust the voltage, current and power parameters of the electrical equipment through the preset parameter adjustment rule, and record the adjusted equipment operation data for subsequent model optimization. This control strategy can adjust reasonably according to the different states of the equipment and the power grid, and ensure the safe operation of the equipment.

[0055] Control execution and feedback optimization: The control strategy is sent to the electrical equipment actuator through the control module, and the control effect is monitored in real time and optimized. When monitoring the control effect in real time, the effectiveness of the control strategy is judged by comparing the change amplitude of the electrical equipment operating parameters before and after control, and whether the state recognition result returns to normal. If the control effect does not meet the expectation, the control strategy is adjusted according to the preset feedback optimization rule and executed again. Through the feedback optimization mechanism, the control strategy is continuously improved, and the control effect is improved.

[0056] Based on the above method, the application also provides an intelligent state recognition and control system for electrical equipment, comprising:

[0057] Data acquisition module: including current sensor, voltage sensor, temperature sensor, vibration sensor, humidity sensor and network communication module, used for real-time acquisition of electrical equipment operation data, historical data, maintenance records and power grid load information.

[0058] Data preprocessing module: filtering and normalizing the collected raw data to improve data quality.

[0059] Feature extraction module: based on convolutional neural network, the preprocessed data is feature extracted to generate key feature vector.

[0060] State recognition module: using support vector machine model, the running state of electrical equipment is judged according to the extracted feature vector.

[0061] Control strategy generation module: generating corresponding control strategy according to device state, power grid load and operating parameters.

[0062] Control execution and feedback module: sending the control strategy to the actuator, and monitoring the control effect in real time, and optimizing according to the feedback.

[0063] The specific implementation process of the intelligent state recognition and control method and system for electrical equipment is as follows:

[0064] Data acquisition: installing current sensor, voltage sensor, temperature sensor, vibration sensor and humidity sensor on the transformer, and collecting real-time current, voltage, temperature, vibration frequency and environmental humidity data of the transformer; at the same time, obtaining the historical operation data, maintenance records of the transformer and the load information of the power grid through the network communication module.

[0065] Data preprocessing: filtering the collected raw data to remove noise interference; using normalization method to standardize different types and magnitudes of data, so that the data have comparability.

[0066] Feature extraction: The pre-processed multi-dimensional time series data is divided into three-dimensional tensor input convolutional neural network according to time window. The first layer of one-dimensional convolution layer adopts 16 convolution kernels with a size of 3 and a step of 1, and the activation function is ReLU, which extracts short-term time series features; after the maximum pooling layer, the feature dimension is reduced and the noise is suppressed; the second layer of one-dimensional convolution layer adopts 32 convolution kernels with a size of 5 and a step of 1, and the activation function is LeakyReLU, which extracts medium-term time series dependence relationship; then the time series dimension is compressed into a feature vector through the global average pooling layer; batch normalization layer and Dropout layer with a Dropout rate of 0.2 are inserted between the convolution layers to prevent overfitting and accelerate training; finally, the feature is mapped to a low-dimensional space through the full connection layer to generate a 64-dimensional key feature vector. Using the training data containing normal, abnormal and fault state samples, the network is trained using a triplet loss function.

[0067] State recognition: The extracted 64-dimensional key feature vector is input into the multi-class state recognition model based on the SVM-RBF kernel function, and the One-vs-One strategy is used to convert the four-class problem into six binary sub-problems. The key feature vector is standardized by Z-score, and the Mahalanobis distance to the center of the four states is calculated respectively, and the running state of the transformer is determined through the voting mechanism. When there is a fuzzy state, the corresponding determination method is used for processing; when 5 consecutive samples are determined as unknown state, the incremental learning process is started to update the training set.

[0068] Regulation strategy generation: If the transformer is in a slight abnormal state, according to the parameter distribution of the normal state in its historical running data, combined with the current power grid load information, the voltage, current and power parameters of the transformer are adjusted through the preset parameter adjustment rule, and the adjusted running data is recorded; if it is in a serious abnormal or fault state, an alarm is sent and the power supply is cut off according to the preset protection logic.

[0069] Regulation execution and feedback optimization: The control module sends the regulation strategy to the actuator of the transformer, and monitors the change amplitude of the transformer running parameters before and after regulation and the state recognition result in real time. If the regulation effect does not meet the expectation, the regulation strategy is adjusted according to the preset feedback optimization rule and executed again.

[0070] The above describes the present application and its embodiments, which are not limited, and the embodiments shown in the drawings are only one of the embodiments of the present application, and the actual structure is not limited thereto. In summary, if a person skilled in the art is inspired thereby, without departing from the purpose of the present application, similar structure and embodiments can be designed without creative design, which shall belong to the protection scope of the present application.

Claims

1. A method for intelligent state identification and control of electrical equipment, characterized by: The method comprises the following steps: S1, using current sensor, voltage sensor, temperature sensor, vibration sensor and humidity sensor, real-time acquisition of current, voltage, temperature, vibration frequency and environmental humidity data of electrical equipment operation, and simultaneously obtaining device historical operation data, maintenance records and power grid load information through a network communication module; S2, filtering the collected raw data to remove noise, and using a normalization method to standardize different types and magnitudes of data; S3, using a convolutional neural network to extract features from the preprocessed data to obtain a key feature vector reflecting the operating state of the electrical equipment; S4, inputting the extracted feature vector into a state recognition model based on a support vector machine and trained by a large amount of historical data to determine whether the electrical equipment is in a normal operation, slight abnormality, serious abnormality or fault state; S5, generating a control strategy according to the device state recognition result, combining the power grid load information and the device operating parameters: adjusting the operating parameters when the device is in a slight abnormality, and issuing an alarm and cutting off the power supply according to the preset protection logic when the device is in a serious abnormality or fault; S6, sending the control strategy to the electrical equipment actuator through the control module, and monitoring the control effect in real time and feeding back the optimization. 2.The intelligent state recognition and regulation method of an electrical device according to claim 1, characterized in that: The preprocessed multi-dimensional time series data is divided into a three-dimensional tensor input according to a time window, and the tensor dimension is [time step x sensor type number x feature dimension]; A multi-layer convolutional neural network architecture is constructed, which comprises in turn: A first one-dimensional convolutional layer, which uses 16 convolutional kernels with a size of 3, a step of 1, and a ReLU activation function, is used to extract short-term time series features; A max pooling layer with a pooling window size of 2 and a step of 2 is used for feature dimension reduction and noise suppression; A second one-dimensional convolutional layer, which uses 32 convolutional kernels with a size of 5, a step of 1, and a LeakyReLU activation function, is used to extract medium-term time series dependence relationships; A global average pooling layer is used to compress the time series dimension into a feature vector; Batch normalization layers and Dropout layers are inserted between the convolutional layers, with a Dropout rate of 0.2, to prevent overfitting and accelerate training; A fully connected layer is used to map the features to a low-dimensional space to generate a 64-dimensional key feature vector, which contains time-frequency domain features, multi-sensor correlation features and environmental influence features of the electrical equipment operating state; A triplet loss function is used to train the network to enhance the aggregation of similar state features and the separation of different state features, and the training data includes normal, abnormal and fault state samples. 3.The method of claim 1, wherein: The specific method for determining the operating state of the electrical equipment is as follows: The 64-dimensional key feature vector extracted by the convolutional neural network is used as input, and the normal, slight abnormality, serious abnormality and fault states of the device are used as output labels; An SVM-RBF kernel function is used to construct a multi-class state recognition model, and a One-vs-One strategy is used to convert the four-class problem into six two-class sub-problems, each of which is responsible for distinguishing between two states; The key feature vectors are Z-score standardized, and the Mahalanobis distances of the key feature vectors to the centers of the four states are calculated respectively; the final state category is determined through a voting mechanism: six binary classifiers are used for voting, if the number of votes for a certain state exceeds the threshold, and the Mahalanobis distance of the sample to the state is the smallest, it is determined that the device is in the state; when the difference between the votes of any two states is less than 5%, the fuzzy state determination is triggered; The method for fuzzy state determination is: calculating the distance of the key feature vector to each classification hyperplane, if all are greater than the threshold, it is determined as unknown state, otherwise, the two closest categories are selected, and secondary determination is performed combined with the historical state transition probability of the device; When 5 consecutive samples are determined as unknown state, the incremental learning process is started; The most representative 20 samples are selected through active learning for manual labeling, and the training set is updated; a sliding window mechanism is used, and every time 100 labeled samples are added, the earliest 50 samples are discarded to keep the size of the training set stable.

4. The intelligent state recognition and regulation method of an electrical device according to claim 3, characterized in that: In the S5 step, when the device is in a slight abnormal state, the specific way of adjusting the operating parameters is: according to the parameter distribution of the device in the normal state in the historical operation data, combined with the current power grid load information, the voltage, current and power parameters of the electrical equipment are adjusted through the preset parameter adjustment rule, and the adjusted device operation data is recorded for subsequent model optimization.

5. The intelligent state recognition and regulation method of an electrical device according to claim 4, characterized in that: In the S6 step, when the real-time monitoring and control effect is compared, the change amplitude of the electrical equipment operating parameters before and after the control, and whether the state recognition result returns to normal are compared to judge the effectiveness of the control strategy; If the control effect does not meet the expectation, the control strategy is adjusted according to the preset feedback optimization rule and executed again.

6. An intelligent state recognition and control system for electrical equipment, which is used to implement the intelligent state recognition and control method for electrical equipment according to any one of claims 1-5, comprising: a data acquisition module: including a current sensor, a voltage sensor, a temperature sensor, a vibration sensor, a humidity sensor and a network communication module, which is used to acquire real-time electrical equipment operation data, historical data, maintenance records and power grid load information; a data preprocessing module: which is used to filter and normalize the collected raw data; a feature extraction module: which is used to extract features from the preprocessed data based on a convolutional neural network to generate key feature vectors; a state recognition module: which is used to determine the operating state of the electrical equipment according to the extracted key feature vectors by using a support vector machine model; a control strategy generation module: which is used to generate corresponding control strategies according to the device state, power grid load and operating parameters; a control execution and feedback module: which is used to send the control strategies to the execution mechanism, and monitor the control effect in real time, and optimize according to the feedback.

Images