Distribution network overhead line partial discharge monitoring system and method

Abstract

The invention discloses a distribution network overhead line partial discharge monitoring system and a distribution network overhead line partial discharge monitoring method, and relates to the technical field of power system monitoring, the system comprises a sensor module, a data acquisition module, a signal processing module and a central control module, and the monitoring accuracy and reliability are improved by fusing a plurality of innovative algorithms; the time-frequency analysis algorithm captures signal time-frequency features, the fuzzy logic algorithm processes uncertainty, the deep belief network algorithm extracts deep features, and the false alarm rate and the missing report rate are reduced by reasonably determining weight coefficients and fusing the results of the three algorithms. Signal information is enriched, various environments are adapted, the contribution degree of each sensor is dynamically adjusted by determining a weight coefficient, the anti-interference capability is improved, and signals are comprehensively and accurately collected.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring, and in particular to a distribution network overhead line partial discharge monitoring system and method. Background Art

[0002] In the current field of partial discharge monitoring of distribution network overhead lines, traditional monitoring systems have many obvious shortcomings.

[0003] On the one hand, most existing monitoring systems use a single monitoring technology or algorithm. For example, some systems rely only on traditional time-frequency analysis algorithms to extract the characteristics of partial discharge signals. However, this method is often difficult to accurately capture the subtle characteristics of partial discharge in the face of complex distribution network overhead line environments, and is prone to false alarms or missed alarms. Some systems also use a single sensor for signal acquisition, such as using only a high-frequency current sensor. The signals collected are relatively simple and cannot fully reflect the actual situation of partial discharge.

[0004] On the other hand, the existing system is not capable of signal fusion processing. When using multiple sensors at the same time, there is a lack of effective fusion algorithms to integrate the advantages of each sensor, which makes it difficult for the signals collected by different sensors to work together and provide strong support for the accurate judgment of partial discharge. At the same time, the existing monitoring system performs poorly in adapting to different environments. Whether in cities, industrial parks or remote mountainous areas, it is difficult to flexibly adjust parameters and algorithms according to actual conditions, and it is impossible to ensure stable and accurate partial discharge monitoring under various complex conditions. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and provide a distribution network overhead line partial discharge monitoring system and method, which can improve the monitoring accuracy and reliability by integrating multiple innovative algorithms. The time-frequency analysis algorithm captures the time-frequency characteristics of the signal, the fuzzy logic algorithm handles uncertainty, and the deep belief network algorithm extracts deep features. The three algorithm results are integrated by reasonably determining the weight coefficient to reduce the false alarm rate and the missed alarm rate. The sensor module adopts a combination of ultra-high frequency, high-frequency current and ultrasonic sensors to enrich signal information and adapt to various environments. The contribution of each sensor is dynamically adjusted by determining the weight coefficient, the anti-interference ability is improved, and the signal is collected comprehensively and accurately.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: On the one hand, a partial discharge monitoring system for overhead lines of a distribution network, the system comprises a sensor module, a data acquisition module, a signal processing module and a central control module;

[0007] The sensor module is used to collect partial discharge signals of the distribution network overhead lines, the data acquisition module is connected to the sensor module, and performs analog-to-digital conversion on the collected signals, and the signal processing module includes a fusion algorithm unit;

[0008] The fusion algorithm unit uses a fusion algorithm to process the converted signal. The fusion algorithm formula is: M(X)=γ×J(X)+δ×K(X)+∈×L(X), where X is the input signal, J(X) is a feature extraction algorithm based on time-frequency analysis, and the time-frequency feature is obtained by performing short-time Fourier transform on the signal. K(X) is a feature extraction algorithm based on fuzzy logic, and the signal is fuzzified using fuzzy rules to extract features. L(X) is a feature extraction algorithm based on a deep belief network, and deep features are extracted through unsupervised learning. γ, δ, ∈ are weight coefficients, and the determination method thereof is as follows: First, a large number of known partial discharge signal samples are selected, and they are processed separately using J(X), K(X), and L(X) algorithms to obtain their respective accuracy rates P J , P K , P L ,but The central control module is connected to the signal processing module, and determines the partial discharge situation according to the processed signal and issues an alarm.

[0009] Furthermore, the sensor module includes a UHF sensor, a high frequency current sensor and an ultrasonic sensor. The UHF sensor is used to detect the UHF signal generated by partial discharge, and its output signal is recorded as S U HF, high frequency current sensor is used to detect the high frequency current signal generated by partial discharge, and the output signal is S H F, ultrasonic sensor is used to detect ultrasonic signal generated by partial discharge, the output signal is S U S, the three sensors cooperate with each other to collect partial discharge signals more comprehensively. The comprehensive signal S total =a×S U HF+b×S H F+c×S U S, where a, b, and c are weight coefficients, which are determined by evaluating the performance of different types of sensors in different partial discharge scenarios. In a specific scenario, the data collected by each sensor is analyzed separately to calculate its accuracy in partial discharge judgment, which is recorded as P UHF , P HF , P US ,but

[0010]

[0011] Furthermore, the data acquisition module includes a high-precision analog-to-digital converter and a data cache unit. The conversion accuracy of the high-precision analog-to-digital converter is n bits, where the value of n is determined according to the dynamic range and accuracy requirements of the signal. The storage capacity of the data cache unit is M bytes, and the size of M is determined according to the data acquisition rate and processing time requirements of the system. The data acquisition module performs real-time acquisition and conversion on the signal output by the sensor module to ensure the accuracy and integrity of the signal.

[0012] Furthermore, the time-frequency analysis algorithm J(X) of the signal processing module is specifically as follows: short-time Fourier transform is performed on the input signal X to obtain a time-frequency matrix T, and then the eigenvector F in the time-frequency matrix is ​​extracted. j , let the time series of signal X be {x 1 ,x 2 ,…,x N}, the window function is w(t), then the short-time Fourier transform formula is Where t represents time and f represents frequency. By performing short-time Fourier transform on the signals at different time points, the time-frequency matrix T can be obtained, and then the eigenvector F can be extracted. j , in the fusion algorithm, the feature vector F j The feature vectors extracted by K(X) and L(X) are used together in the calculation. Specifically, F j After normalization, it is multiplied by the weight coefficient γ and then weighted summed with the feature vectors extracted by K(X) and L(X) that have been processed accordingly to obtain a more accurate representation of the partial discharge features.

[0013] Furthermore, in the feature extraction algorithm K(X) based on fuzzy logic, fuzzy sets and fuzzy rules are first defined, and the characteristic parameters of the input signal X are set to p 1 ,p 2 ,…,p m , define the fuzzy set in is the fuzzy membership function, and the fuzzy rule is R j :If p 1 Yes A i1 And p 2 Yes A i2 and Yes A im , then the output is B j , the fuzzy set is obtained through fuzzy reasoning, and then the eigenvector F is obtained by defuzzification. K , in the fusion algorithm, F K After normalization, multiply it by the weight coefficient δ, and jThe feature vector extracted by L(X) works together to provide the fusion algorithm with feature information based on fuzzy logic, thus enhancing the analysis capability of partial discharge signals.

[0014] Furthermore, the feature extraction algorithm L(X) of the deep belief network is composed of multiple restricted Boltzmann machines (RBMs) stacked together. Let the input signal be X and the visible layer of the first RBM be v 1 , the hidden layer is 1 , and its energy function is where a i , b j are the biases of the visible layer and the hidden layer, respectively, w ij To connect the weights, train the RBM through the contrast divergence algorithm to obtain the feature representation of the hidden layer. The hidden layer of the first RBM is used as the visible layer of the second RBM. Multiple RBMs are stacked in sequence. Finally, the whole network is fine-tuned through the back propagation algorithm to obtain the deep feature vector F of the input signal. L , in the fusion algorithm, F L After normalization, it is multiplied by the weight coefficient ∈ and F j and F K Together they form a fused feature vector, providing more comprehensive and accurate information for the central control module to judge the partial discharge situation.

[0015] Furthermore, the central control module includes a high-performance microprocessor and a communication unit. The processing frequency of the microprocessor is f Hz, and the size of f is determined according to the real-time requirements of the system. The communication unit adopts wireless communication technology, including but not limited to 5G communication module, and its data transmission rate is k bits / second. The value of k is determined according to the data transmission volume and real-time requirements of the system. The central control module receives the signal output by the signal processing module, analyzes and judges, and then issues an alarm signal.

[0016] Furthermore, the monitoring system further comprises a power module, which is powered by a combination of solar cells and lithium batteries. Assuming that the output power of the solar cell is P solar The capacity of the lithium battery is C ampere-hours. During the day, the solar cell powers the system and charges the lithium battery. At night or on cloudy days, the lithium battery powers the system. The power management unit performs intelligent management based on the system's power consumption and battery status to ensure stable operation of the system.

[0017] On the other hand, a method for using a partial discharge monitoring system for a distribution network overhead line is provided, wherein the specific steps of the method are as follows:

[0018] Step 1: Install the sensor module. Select a suitable location on the overhead line of the distribution network for installation. Ensure that the UHF sensor, high-frequency current sensor, and ultrasonic sensor can stably contact the line and are not interfered by the outside world. During the installation process, strictly follow the installation instructions to ensure that the sensor is installed firmly and reliably.

[0019] Step 2: Start the data acquisition module, turn on the data acquisition module, and establish a connection with the sensor module. The data acquisition module will collect the analog signal output by the sensor in real time, and perform analog-to-digital conversion to convert it into a digital signal for subsequent processing;

[0020] Step 3: Signal processing: The signal processing module receives the digital signal output by the data acquisition module. First, it uses the feature extraction algorithm J(X) based on time-frequency analysis to perform short-time Fourier transform on the input signal to obtain the time-frequency matrix T and extract the feature vector F. j Then, we use the feature extraction algorithm K(X) based on fuzzy logic to define fuzzy sets and fuzzy rules, perform fuzzy processing and reasoning on the signal, and obtain the feature vector F K Then, the feature extraction algorithm L(X) based on deep belief network is used to extract the deep feature vector F by stacking multiple restricted Boltzmann machines for unsupervised learning. L ,Finally, according to the fusion algorithm formula M(X)=γ×J(X)+δ×K(X)+∈×L(X), the three feature vectors are weighted fused to obtain the partial discharge feature representation;

[0021] Step 4: Judgment and alarm. The central control module receives the fused feature representation output by the signal processing module and uses the preset judgment algorithm to analyze it to determine whether there is partial discharge in the distribution network overhead line. If the judgment result is that there is partial discharge, the central control module sends an alarm signal to notify relevant personnel to handle it.

[0022] Step 5: Regular maintenance. During system operation, regularly calibrate and maintain the sensors, check whether the sensor connections are loose or damaged. At the same time, perform performance tests on the data acquisition module, signal processing module and central control module to ensure that all parts of the system are operating normally. According to actual conditions, regularly update system parameters and algorithms to improve monitoring accuracy and reliability.

[0023] Compared with the prior art, the distribution network overhead line partial discharge monitoring system and method have the following beneficial effects:

[0024] 1. The distribution network overhead line partial discharge monitoring system of the present invention significantly improves the accuracy and reliability of monitoring by integrating multiple innovative algorithms. The time-frequency analysis algorithm J(X), the fuzzy logic algorithm K(X) and the deep belief network algorithm L(X) in the fusion algorithm respectively extract the characteristics of the partial discharge signal from different angles, giving full play to their respective advantages. Among them, the time-frequency analysis algorithm can accurately capture the time-frequency characteristics of the signal, the fuzzy logic algorithm can handle the uncertainty in the signal, and the deep belief network algorithm can extract deep features. By reasonably determining the weight coefficient, the results of the three algorithms are integrated, making the judgment of partial discharge more accurate. Compared with the traditional single algorithm monitoring system, the present invention greatly reduces the false alarm rate and the missed alarm rate, and provides a strong guarantee for the safe operation of the distribution network overhead lines.

[0025] Second, the sensor module of the present invention adopts a combination of ultra-high frequency sensor, high frequency current sensor and ultrasonic sensor, which greatly enriches the collected partial discharge signal information. The three sensors have their own advantages in different partial discharge scenarios. Through the comprehensive signal formula S total =a×S U HF+b×S H F+c×S U S determines the weight coefficient and can dynamically adjust the contribution of each sensor according to the actual situation. Such a design can adapt to various complex distribution network overhead line environments. Whether in cities, industrial parks or remote mountainous areas, it can comprehensively and accurately collect local discharge signals. At the same time, the collaborative work of different types of sensors also improves the system's anti-interference ability, effectively avoiding the interference factors that a single sensor may be affected by, and further improving the performance of the monitoring system.

[0026] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 It is a flow chart of a partial discharge monitoring system for overhead lines in a distribution network;

[0029] Figure 2The present invention is a flow chart of a method for using a partial discharge monitoring system for overhead lines in a distribution network. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. 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.

[0031] Embodiment 1:

[0032] The monitoring system of the present invention is adopted in a partial discharge monitoring project of overhead lines of a distribution network in a large industrial city.

[0033] First, the sensor module is installed. According to the layout of the urban distribution network overhead lines and potential local discharge risk points, ultra-high frequency sensors, high-frequency current sensors and ultrasonic sensors are installed on important substation outlet towers, load-concentrated area towers and old line towers. The ultra-high frequency sensor adopts a high-performance antenna design and can sensitively detect the ultra-high frequency signal generated by local discharge. The high-frequency current sensor uses a Rogowski coil structure to accurately capture the high-frequency current changes caused by local discharge. The ultrasonic sensor uses piezoelectric ceramic materials to quickly respond to the ultrasonic signals generated by local discharge. During the installation process, strictly follow the installation specifications to ensure that the distance between the sensor and the line is appropriate and firmly fixed to prevent the sensor position from shifting due to wind, vibration and other factors.

[0034] For the data acquisition module, a high-precision analog-to-digital converter with a conversion accuracy of 16 bits is selected, which can accurately convert the analog signals collected by the sensor into digital signals. The storage capacity of the data cache unit is 512MB. When there is a brief interruption in data transmission or the signal processing module is busy, data for a certain period of time can be stored to avoid data loss. The data acquisition module operates at a higher sampling frequency to ensure that subtle changes in partial discharge signals can be captured.

[0035] In the signal processing module, in order to determine the weight coefficient of the fusion algorithm, a large number of known partial discharge signal samples were analyzed. After repeated testing and verification, the accuracy of the feature extraction algorithm J(X) based on time-frequency analysis in these samples was P J The algorithm can effectively extract the time-frequency characteristics of the signal by performing short-time Fourier transform on the input signal, thus providing an important basis for the judgment of partial discharge. The accuracy of the feature extraction algorithm K(X) based on fuzzy logic is P KThe algorithm can handle the uncertainty and ambiguity in the signal and improve the accuracy of partial discharge feature extraction by defining fuzzy sets and fuzzy rules. The accuracy of feature extraction algorithm L(X) based on deep belief network is P L The algorithm performs unsupervised learning by stacking multiple restricted Boltzmann machines, which can extract deep features of the signal and has strong recognition ability for complex partial discharge signals.

[0036] According to the method of determining the weight coefficient,

[0037] For the ultra-high frequency signal S collected by the sensor module U HF, high frequency current signal S H F and ultrasonic signal S U After the performance evaluation in the current urban distribution network environment, the accuracy of the three sensors for partial discharge judgment is P UHF =70%,P HF =65%, P US =75%, then

[0038] The monitoring process is as follows: the UHF sensor, high-frequency current sensor and ultrasonic sensor work simultaneously to collect the UHF signal, high-frequency current signal and ultrasonic signal generated by partial discharge respectively. These signals are transmitted to the data acquisition module, and are stored in the data cache unit after analog-to-digital conversion. Then, the signal enters the signal processing module. First, the signal is subjected to short-time Fourier transform based on the feature extraction algorithm J(X) of time-frequency analysis to obtain the time-frequency matrix and extract the feature vector F j Then, based on the fuzzy logic feature extraction algorithm K(X), fuzzy sets and fuzzy rules are defined, and the signal is fuzzified and inferred to obtain the feature vector F K Then, the feature extraction algorithm L(X) based on the deep belief network performs unsupervised learning by stacking multiple restricted Boltzmann machines to extract the deep feature vector F L Finally, according to the fusion algorithm formula M(X)=γ×J(X)+δ×K(X)+∈×L(X), the three feature vectors are weighted fused to obtain the partial discharge feature representation. After receiving this feature representation, the central control module uses the preset judgment algorithm to analyze and judge whether there is partial discharge in the distribution network overhead line. If the judgment result is that there is partial discharge, an alarm signal is issued to notify relevant personnel to handle it.

[0039] Through the coordinated work of the three sensors, local discharge signals can be fully collected, improving the accuracy of monitoring. The application of fusion algorithms gives full play to the advantages of different feature extraction algorithms, making the judgment of local discharge more accurate and reliable. The rapid response and alarm function of the central control module can promptly notify relevant personnel for processing, effectively avoiding power failures that may be caused by local discharge and ensuring the safe and stable operation of the urban distribution network.

[0040] Embodiment 2:

[0041] The system of the present invention is applied in the monitoring of overhead lines of a distribution network in a modern industrial park.

[0042] The installation locations of the sensor modules have been carefully selected. They are mainly installed at key line nodes in the park, incoming poles of large power users, and line sections that are susceptible to external interference. The UHF sensor adopts advanced filtering technology, which can effectively eliminate interference from other wireless signals in the park and accurately detect the UHF signals generated by partial discharge. The high-frequency current sensor has high sensitivity and wide dynamic range, which is suitable for partial discharge detection under different load conditions in the park. The ultrasonic sensor adopts waterproof and dustproof design and can work stably in harsh industrial environments.

[0043] The data acquisition module uses a 14-bit analog-to-digital converter, which reduces system costs while ensuring data accuracy. The data cache unit capacity is 256MB, which can meet the relatively high data acquisition frequency and data storage requirements within a certain period of time in the park. A reliable communication protocol is used between the data acquisition module and the sensor module to ensure the stability and accuracy of data transmission.

[0044] In the signal processing module, a large number of partial discharge signal samples collected in the industrial park are analyzed to obtain the accuracy P of the feature extraction algorithm J(X) based on time-frequency analysis. J The algorithm can effectively extract the time-frequency characteristics of partial discharge signals in the complex electromagnetic environment of industrial parks, providing an accurate data basis for subsequent analysis. The accuracy of the feature extraction algorithm K(X) based on fuzzy logic is P K The algorithm can process the uncertainty signals generated by the operation of various electrical equipment in the industrial park and improve the accuracy of partial discharge feature extraction. The accuracy of the feature extraction algorithm L(X) based on the deep belief network is P L The algorithm can automatically adapt to the local discharge signal characteristics under different working conditions in industrial parks through deep learning, and improve the accuracy and reliability of monitoring.

[0045] The weight coefficient is calculated from this

[0046] For sensor signals, the accuracy of UHF, HF current and ultrasonic sensors in this scenario are P UHF =68%,P HF =63%, P US =73%, so the weight coefficient

[0047] The monitoring process is as follows: three sensors collect signals in real time, the data acquisition module converts analog signals into digital signals and stores them, and after the signals enter the signal processing module, the J(X) algorithm extracts the time-frequency characteristics of the signals, the K(X) algorithm performs fuzzy processing and reasoning, and the L(X) algorithm extracts deep features. Then, a comprehensive partial discharge feature representation is obtained through the fusion algorithm. The central control module determines whether there is partial discharge based on this feature representation and issues an alarm signal when necessary.

[0048] In the complex electromagnetic environment of the industrial park, the system can accurately monitor partial discharge. The filtering technology of the UHF sensor effectively avoids interference and ensures the accuracy of the signal. The fusion algorithm combines the advantages of three different algorithms and can adapt to various working conditions in the industrial park, thereby improving the reliability of monitoring. The timely alarm function allows park managers to take quick measures to prevent power accidents caused by partial discharge, thereby ensuring production safety and stable power supply in the industrial park.

[0049] Embodiment three:

[0050] The system of the present invention is used in partial discharge monitoring of overhead lines of a distribution network in a remote mountainous area.

[0051] Due to the complex environment, inconvenient transportation and difficult signal transmission in mountainous areas, higher requirements are placed on the stability and reliability of the system. Sensor modules are installed at locations on mountain lines where partial discharge is prone to occur, such as poles and towers across valleys, line sections close to trees, and old line poles and towers. UHF sensors are designed with low power consumption to adapt to the limited power supply conditions in mountainous areas. High-frequency current sensors have good anti-interference capabilities and can accurately detect partial discharge signals in complex electromagnetic environments in mountainous areas. Ultrasonic sensors use highly sensitive sensor elements that can capture weak partial discharge ultrasonic signals.

[0052] The data acquisition module uses a 12-bit analog-to-digital converter, which reduces the power consumption of the system while meeting the data accuracy requirements. The data cache unit has a capacity of 128MB. Considering that the communication conditions in mountainous areas may be poor, the data cache unit can store data for a certain period of time when data transmission is interrupted, and then transmit it after the communication is restored. Wireless communication is used between the data acquisition module and the sensor module, reducing the difficulty and cost of wiring in mountainous areas.

[0053] In the signal processing module, by analyzing the local discharge signal samples collected in the mountainous area, the accuracy of the J(X) algorithm P J is 75%, the accuracy of K(X) algorithm is P K is 70%, the accuracy of L(X) algorithm P L The weight coefficient is 80%.

[0054]

[0055] For sensor signals, the accuracy rates of UHF, HF current and ultrasonic sensors in this mountain scene are P UHF =65%,P HF =60%, P US =70%, weight coefficient

[0056] The monitoring process is as follows: the sensor module continuously collects signals, the data acquisition module performs analog-to-digital conversion and storage, and after the signal enters the signal processing module, it is processed by three feature extraction algorithms respectively, and then the local discharge feature representation is obtained through the fusion algorithm. The central control module determines whether there is local discharge and sends out an alarm signal when necessary. Due to the poor communication conditions in mountainous areas, the communication unit adopts a combination of satellite communication and ground communication to ensure that data can be transmitted stably.

[0057] In remote mountainous areas, the system has successfully realized partial discharge monitoring of distribution network overhead lines. The low-power sensor design and data acquisition module adapt to the power supply conditions in mountainous areas. The application of fusion algorithms improves the accuracy of monitoring and can work reliably even in complex mountainous environments. The combination of satellite communication and ground communication ensures the stability of data transmission, allowing the remote monitoring center to promptly understand the operation of mountain lines. The timely alarm function provides a guarantee for the power supply in mountainous areas, ensuring that the production and life of mountain residents are not affected by power failures.

[0058] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A partial discharge monitoring system for overhead lines in a distribution network, characterized in that: The system includes a sensor module, a data acquisition module, a signal processing module and a central control module; The sensor module is used to collect partial discharge signals of the distribution network overhead lines, the data acquisition module is connected to the sensor module, and performs analog-to-digital conversion on the collected signals, and the signal processing module includes a fusion algorithm unit; The fusion algorithm unit uses a fusion algorithm to process the converted signal. The fusion algorithm formula is: M(X)=γ×J(X)+δ×K(X)+∈×L(X), where X is the input signal, J(X) is a feature extraction algorithm based on time-frequency analysis, and the time-frequency feature is obtained by performing short-time Fourier transform on the signal. K(X) is a feature extraction algorithm based on fuzzy logic, and the signal is fuzzified using fuzzy rules to extract features. L(X) is a feature extraction algorithm based on a deep belief network, and deep features are extracted through unsupervised learning. γ, δ, ∈ are weight coefficients, and the determination method thereof is as follows: First, a large number of known partial discharge signal samples are selected, and they are processed separately using J(X), K(X), and L(X) algorithms to obtain their respective accuracy rates P J , P K , P L ,but The central control module is connected to the signal processing module, and determines the partial discharge situation according to the processed signal and issues an alarm.

2. A distribution network overhead line partial discharge monitoring system according to claim 1, characterized in that: The sensor module includes a UHF sensor, a high-frequency current sensor and an ultrasonic sensor. The UHF sensor is used to detect the UHF signal generated by partial discharge, and its output signal is recorded as S U HF, high frequency current sensor is used to detect the high frequency current signal generated by partial discharge, and the output signal is S H F, ultrasonic sensor is used to detect ultrasonic signal generated by partial discharge, the output signal is S U S, the three sensors cooperate with each other to collect partial discharge signals more comprehensively. The comprehensive signal S total =a×S U HF+b×S H F+c×S U S, where a, b, and c are weight coefficients, which are determined by evaluating the performance of different types of sensors in different partial discharge scenarios. In a specific scenario, the data collected by each sensor is analyzed separately to calculate its accuracy in partial discharge judgment, which is recorded as P UHF , P HF , P US ,but 3. A distribution network overhead line partial discharge monitoring system according to claim 1, characterized in that: The data acquisition module includes a high-precision analog-to-digital converter and a data cache unit. The conversion accuracy of the high-precision analog-to-digital converter is n bits, where the value of n is determined according to the dynamic range and accuracy requirements of the signal. The storage capacity of the data cache unit is M bytes, and the size of M is determined according to the data acquisition rate and processing time requirements of the system. The data acquisition module collects and converts the signal output by the sensor module in real time to ensure the accuracy and integrity of the signal.

4. A distribution network overhead line partial discharge monitoring system according to claim 1, characterized in that: The time-frequency analysis algorithm J(X) of the signal processing module is as follows: short-time Fourier transform is performed on the input signal X to obtain a time-frequency matrix T, and then the eigenvector F in the time-frequency matrix is ​​extracted. j , let the time series of signal X be {x1,x2,…,x N }, the window function is w(t), then the short-time Fourier transform formula is Where t represents time and f represents frequency. By performing short-time Fourier transform on the signals at different time points, the time-frequency matrix T can be obtained, and then the eigenvector F can be extracted. j , in the fusion algorithm, the feature vector F j The feature vectors extracted by K(X) and L(X) are used together in the calculation. Specifically, F j After normalization, it is multiplied by the weight coefficient γ and then weighted summed with the feature vectors extracted by K(X) and L(X) that have been processed accordingly to obtain a more accurate representation of the partial discharge features.

5. A distribution network overhead line partial discharge monitoring system according to claim 4, characterized in that: In the feature extraction algorithm K(X) based on fuzzy logic, fuzzy sets and fuzzy rules are first defined. Suppose the characteristic parameters of the input signal X are p1, p2, …, p m , define the fuzzy set in is the fuzzy membership function, and the fuzzy rule is R j :If p1 is A i1 And p2 is A i2 and Yes A im , then the output is B j , the fuzzy set is obtained through fuzzy reasoning, and then the eigenvector F is obtained by defuzzification. K , in the fusion algorithm, F K After normalization, multiply it by the weight coefficient δ, and j The feature vector extracted by L(X) works together to provide the fusion algorithm with feature information based on fuzzy logic, thus enhancing the analysis capability of partial discharge signals.

6. A distribution network overhead line partial discharge monitoring system according to claim 5, characterized in that: The feature extraction algorithm L(X) of the deep belief network is composed of multiple restricted Boltzmann machines (RBMs) stacked together. Let the input signal be X, the visible layer of the first RBM be v1, and the hidden layer be 1. Its energy function is where a i , b j are the biases of the visible layer and the hidden layer, respectively, w ij To connect the weights, train the RBM through the contrast divergence algorithm to obtain the feature representation of the hidden layer. The hidden layer of the first RBM is used as the visible layer of the second RBM. Multiple RBMs are stacked in sequence. Finally, the whole network is fine-tuned through the back propagation algorithm to obtain the deep feature vector F of the input signal. L , in the fusion algorithm, F L After normalization, it is multiplied by the weight coefficient ∈ and F j and F K Together they form a fused feature vector, providing more comprehensive and accurate information for the central control module to judge the partial discharge situation.

7. A distribution network overhead line partial discharge monitoring system according to claim 1, characterized in that: The central control module includes a high-performance microprocessor and a communication unit. The processing frequency of the microprocessor is f Hz, and the size of f is determined according to the real-time requirements of the system. The communication unit adopts wireless communication technology, including but not limited to 5G communication module, and its data transmission rate is k bits / second. The value of k is determined according to the data transmission volume and real-time requirements of the system. The central control module receives the signal output by the signal processing module, analyzes and judges, and then issues an alarm signal.

8. A distribution network overhead line partial discharge monitoring system according to claim 1, characterized in that: The monitoring system also includes a power module, which is powered by a combination of solar cells and lithium batteries. Assume that the output power of the solar cell is P solar The capacity of the lithium battery is C ampere-hours. During the day, the solar cell powers the system and charges the lithium battery. At night or on cloudy days, the lithium battery powers the system. The power management unit performs intelligent management based on the system's power consumption and battery status to ensure stable operation of the system.

9. A method for using a partial discharge monitoring system for overhead lines in a distribution network according to any one of claims 1 to 8, characterized in that: The specific steps of this method are: Step 1: Install the sensor module. Select a suitable location on the distribution network overhead line for installation. Ensure that the UHF sensor, high-frequency current sensor, and ultrasonic sensor can stably contact the line and are not interfered by the outside world. During the installation process, strictly follow the installation instructions to ensure that the sensor is installed firmly and reliably. Step 2: Start the data acquisition module, turn on the data acquisition module, and establish a connection with the sensor module. The data acquisition module will collect the analog signal output by the sensor in real time, and perform analog-to-digital conversion to convert it into a digital signal for subsequent processing; Step 3: Signal processing: The signal processing module receives the digital signal output by the data acquisition module. First, it uses the feature extraction algorithm J(X) based on time-frequency analysis to perform short-time Fourier transform on the input signal to obtain the time-frequency matrix T and extract the feature vector F. j Then, we use the feature extraction algorithm K(X) based on fuzzy logic to define fuzzy sets and fuzzy rules, perform fuzzy processing and reasoning on the signal, and obtain the feature vector F K Then, the feature extraction algorithm L(X) based on deep belief network is used to extract the deep feature vector F by stacking multiple restricted Boltzmann machines for unsupervised learning. L ,Finally, according to the fusion algorithm formula M(X)=γ×J(X)+δ×K(X)+∈×L(X), the three feature vectors are weighted fused to obtain the partial discharge feature representation; Step 4: Judgment and alarm. The central control module receives the fused feature representation output by the signal processing module and uses the preset judgment algorithm to analyze it to determine whether there is partial discharge in the distribution network overhead line. If the judgment result is that there is partial discharge, the central control module sends an alarm signal to notify relevant personnel to handle it. Step 5: Regular maintenance. During system operation, regularly calibrate and maintain the sensors, check whether the sensor connections are loose or damaged. At the same time, perform performance tests on the data acquisition module, signal processing module and central control module to ensure that all parts of the system are operating normally. According to actual conditions, regularly update system parameters and algorithms to improve monitoring accuracy and reliability.

Images